carbon-c-relay(1) -- graphite relay, aggregator and rewriter

SYNOPSIS

carbon-c-relay -f config-file [ options ... ]

DESCRIPTION

carbon-c-relay accepts, cleanses, matches, rewrites, forwards and aggregates graphite metrics by listening for incoming connections and relaying the messages to other servers defined in its configuration. The core functionality is to route messages via flexible rules to the desired destinations.

carbon-c-relay is a simple program that reads its routing information from a file. The command line arguments allow to set the location for this file, as well as the amount of dispatchers (worker threads) to use for reading the data from incoming connections and passing them onto the right destination(s). The route file supports two main constructs: clusters and matches. The first define groups of hosts data metrics can be sent to, the latter define which metrics should be sent to which cluster. Aggregation rules are seen as matches. Rewrites are actions that directly affect the metric at the point in which they appear in the configuration.

For every metric received by the relay, cleansing is performed. The following changes are performed before any match, aggregate or rewrite rule sees the metric:

• double dot elimination (necessary for correctly functioning consistent hash routing)
• trailing/leading dot elimination
• whitespace normalisation (this mostly affects output of the relay to other targets: metric, value and timestamp will be separated by a single space only, ever)
• irregular char replacement with underscores (_), currently irregular is defined as not being in [0-9a-zA-Z-_:#], but can be overridden on the command line. Note that tags (when present and allowed) are not processed this way.

OPTIONS

These options control the behaviour of carbon-c-relay.

• -v: Print version string and exit.

• -d: Enable debug mode, this prints statistics to stdout and prints extra messages about some situations encountered by the relay that normally would be too verbose to be enabled. When combined with -t (test mode) this also prints stub routes and consistent-hash ring contents.

• -s: Enable submission mode. In this mode, internal statistics are not generated. Instead, queue pressure and metrics drops are reported on stdout. This mode is useful when used as submission relay which' job is just to forward to (a set of) main relays. Statistics about the submission relays in this case are not needed, and could easily cause a non-desired flood of metrics e.g. when used on each and every host locally.

• -S: Enable iostat-like mode where every second the current state of statistics are reported. This implies submission mode -s.

• -t: Test mode. This mode doesn't do any routing at all, but instead reads input from stdin and prints what actions would be taken given the loaded configuration. This mode is very useful for testing relay routes for regular expression syntax etc. It also allows to give insight on how routing is applied in complex configurations, for it shows rewrites and aggregates taking place as well. When -t is repeated, the relay will only test the configuration for validity and exit immediately afterwards. Any standard output is suppressed in this mode, making it ideal for start-scripts to test a (new) configuration.

• -f config-file: Read configuration from config-file. A configuration consists of clusters and routes. See CONFIGURATION SYNTAX for more information on the options and syntax of this file.

• -l log-file: Use log-file for writing messages. Without this option, the relay writes both to stdout and stderr. When logging to file, all messages are prefixed with MSG when they were sent to stdout, and ERR when they were sent to stderr.

• -p port: Listen for connections on port port. The port number is used for both TCP, UDP and UNIX sockets. In the latter case, the socket file contains the port number. The port defaults to 2003, which is also used by the original carbon-cache.py. Note that this only applies to the defaults, when listen directives are in the config, this setting is ignored.

• -w workers: Use workers number of threads. The default number of workers is equal to the amount of detected CPU cores. It makes sense to reduce this number on many-core machines, or when the traffic is low.

• -b batchsize: Set the amount of metrics that sent to remote servers at once to batchsize. When the relay sends metrics to servers, it will retrieve batchsize metrics from the pending queue of metrics waiting for that server and send those one by one. The size of the batch will have minimal impact on sending performance, but it controls the amount of lock-contention on the queue. The default is 2500.

• -q queuesize: Each server from the configuration where the relay will send metrics to, has a queue associated with it. This queue allows for disruptions and bursts to be handled. The size of this queue will be set to queuesize which allows for that amount of metrics to be stored in the queue before it overflows, and the relay starts dropping metrics. The larger the queue, more metrics can be absorbed, but also more memory will be used by the relay. The default queue size is 25000.

• -L stalls: Sets the max mount of stalls to stalls before the relay starts dropping metrics for a server. When a queue fills up, the relay uses a mechanism called stalling to signal the client (writing to the relay) of this event. In particular when the client sends a large amount of metrics in very short time (burst), stalling can help to avoid dropping metrics, since the client just needs to slow down for a bit, which in many cases is possible (e.g. when catting a file with nc(1)). However, this behaviour can also obstruct, artificially stalling writers which cannot stop that easily. For this the stalls can be set from 0 to 15, where each stall can take around 1 second on the client. The default value is set to 4, which is aimed at the occasional disruption scenario and max effort to not loose metrics with moderate slowing down of clients.

• -C CAcertpath: Read CA certs (for use with TLS/SSL connections) from given path or file. When not given, the default locations are used. Strict verfication of the peer is performed, so when using self-signed certificates, be sure to include the CA cert in the default location, or provide the path to the cert using this option.

• -T timeout: Specifies the IO timeout in milliseconds used for server connections. The default is 600 milliseconds, but may need increasing when WAN links are used for target servers. A relatively low value for connection timeout allows the relay to quickly establish a server is unreachable, and as such failover strategies to kick in before the queue runs high.

• -c chars: Defines the characters that are next to [A-Za-z0-9] allowed in metrics to chars. Any character not in this list, is replaced by the relay with _ (underscore). The default list of allowed characters is -_:#.

• -m length: Limits the metric names to be of at most length bytes long. Any lines containing metric names larger than this will be discarded.

• -M length Limits the input to lines of at most length bytes. Any excess lines will be discarded. Note that -m needs to be smaller than this value.

• -H hostname: Override hostname determined by a call to gethostname(3) with hostname. The hostname is used mainly in the statistics metrics carbon.relays.<hostname>.<...> sent by the relay.

• -B backlog: Sets TCP connection listen backlog to backlog connections. The default value is 32 but on servers which receive many concurrent connections, this setting likely needs to be increased to avoid connection refused errors on the clients.

• -U bufsize: Sets the socket send/receive buffer sizes in bytes, for both TCP and UDP scenarios. When unset, the OS default is used. The maximum is also determined by the OS. The sizes are set using setsockopt with the flags SO_RCVBUF and SO_SNDBUF. Setting this size may be necessary for large volume scenarios, for which also -B might apply. Checking the Recv-Q and the receive errors values from netstat gives a good hint about buffer usage.

• -E: Disable disconnecting idle incoming connections. By default the relay disconnects idle client connections after 10 minutes. It does this to prevent resources clogging up when a faulty or malicious client keeps on opening connections without closing them. It typically prevents running out of file descriptors. For some scenarios, however, it is not desirable for idle connections to be disconnected, hence passing this flag will disable this behaviour.

• -D: Deamonise into the background after startup. This option requires -l and -P flags to be set as well.

• -P pidfile: Write the pid of the relay process to a file called pidfile. This is in particular useful when daemonised in combination with init managers.

• -O threshold: The minimum number of rules to find before trying to optimise the ruleset. The default is 50, to disable the optimiser, use -1, to always run the optimiser use 0. The optimiser tries to group rules to avoid spending excessive time on matching expressions.

CONFIGURATION SYNTAX

The config file supports the following syntax, where comments start with a # character and can appear at any position on a line and suppress input until the end of that line:

cluster <name>
    < <forward | any_of | failover> [useall] |
      <carbon_ch | fnv1a_ch | jump_fnv1a_ch> [replication <count>] [dynamic] >
        <host[:port][=instance] [proto <udp | tcp>]
                                [type linemode]
                                [transport <plain | gzip | lz4 | snappy>
                                           [ssl | mtls <pemcert> <pemkey>]]> ...
    ;

cluster <name>
    file [ip]
        </path/to/file> ...
    ;

match
        <* | expression ...>
    [validate <expression> else <log | drop>]
    send to <cluster ... | blackhole>
    [stop]
    ;

rewrite <expression>
    into <replacement>
    ;

aggregate
        <expression> ...
    every <interval> seconds
    expire after <expiration> seconds
    [timestamp at <start | middle | end> of bucket]
    compute <sum | count | max | min | average |
             median | percentile<%> | variance | stddev> write to
        <metric>
    [compute ...]
    [send to <cluster ...>]
    [stop]
    ;

send statistics to <cluster ...>
    [stop]
    ;
statistics
    [submit every <interval> seconds]
    [reset counters after interval]
    [prefix with <prefix>]
    [send to <cluster ...>]
    [stop]
    ;

listen
    type linemode [transport <plain | gzip | lz4 | snappy>
                      [<ssl | mtls> <pemcert>
                          [protomin <tlsproto>] [protomax <tlsproto>]
                          [ciphers <ssl-ciphers>] [ciphersuites <tls-suite>]
                      ]
                  ]
        <<interface[:port] | port> proto <udp | tcp>> ...
        </ptah/to/file proto unix> ...
    ;
    # tlsproto: <ssl3 | tls1.0 | tls1.1 | tls1.2 | tls1.3>
    # ssl-ciphers: see ciphers(1)
    # tls-suite: see SSL_CTX_set_ciphersuites(3)

include </path/to/file/or/glob>
    ;

CLUSTERS

Multiple clusters can be defined, and need not to be referenced by a match rule. All clusters point to one or more hosts, except the file cluster which writes to files in the local filesystem. host may be an IPv4 or IPv6 address, or a hostname. Since host is followed by an optional : and port, for IPv6 addresses not to be interpreted wrongly, either a port must be given, or the IPv6 address surrounded by brackets, e.g. [::1]. Optional transport and proto clauses can be used to wrap the connection in a compression or encryption layer or specify the use of UDP or TCP to connect to the remote server. When omitted the connection defaults to a plain TCP connection. type can only be linemode at the moment, e.g. Python's pickle mode is not supported.

DNS hostnames are resolved to a single address, according to the preference rules in RFC 3484. The any_of, failover and forward clusters have an explicit useall flag that enables expansion for hostnames resolving to multiple addresses. Using this option, each address of any type becomes a cluster destination. This means for instance that both IPv4 and IPv6 addresses are added.

There are two groups of cluster types, simple forwarding clusters and consistent hashing clusters.

• forward and file clusters

  The forward and file clusters simply send everything they receive to the defined members (host addresses or files). When a cluster has multiple members, all incoming metrics are sent to all members, basically duplicating the input metric stream over all members.

• any_of cluster

  The any_of cluster is a small variant of the forward cluster, but instead of sending the input metrics to all defined members, it sends each incoming metric to only one of the defined members. The purpose of this is a load-balanced scenario where any of the members can receive any metric. As any_of suggests, when any of the members become unreachable, the remaining available members will immediately receive the full input stream of metrics. This specifically mean that when 4 members are used, each will receive approximately 25% of the input metrics. When one member becomes unavailable (e.g. network interruption, or a restart of the service), the remaining 3 members will each receive about 25% / 3 = ~8% more traffic (33%). Alternatively, they will receive 1/3rd the total input. When designing cluster capacity, one should take into account that in the most extreme case, the final remaining member will receive all input traffic.

  An any_of cluster can in particular be useful when the cluster points to other relays or caches. When used with other relays, it effectively load-balances, and adapts immediately to inavailability of targets. When used with caches, there is a small detail to how any_of works, that makes it very suitable. The implementation of this router is not to round-robin over any available members, but instead it uses a consistent hashing strategy to deliver the same metrics to the same destination all the time. This helps caches, and makes it easier to retrieve uncommitted datapoints (from a single cache), but still allows for a rolling restart of the caches. When a member becomes unavailable, the hash destinations are not changed, but instead traffic destined for the unavailable node is spread evenly over available nodes.

• failover cluster

  The failover cluster is like the any_of cluster, but sticks to the order in which servers are defined. This is to implement a pure failover scenario between servers. All metrics are sent to at most 1 member, so no hashing or balancing is taking place. For example, a failover cluster with two members will only send metrics to the second member when the first member becomes unavailable. As soon as the first member returns, all metrics are sent to the first node again.

• carbon_ch cluster

  The carbon_ch cluster sends the metrics to the member that is responsible according to the consistent hash algorithm as used in the original carbon python relay. Multiple members are possible if replication is set to a value higher than 1. When dynamic is set, failure of any of the servers does not result in metrics being dropped for that server, but instead the undeliverable metrics are sent to any other server in the cluster in order for the metrics not to get lost. This is most useful when replication is 1.

  The calculation of the hashring, that defines the way in which metrics are distributed, is based on the server host (or IP address) and the optional instance of the member. This means that using carbon_ch two targets on different ports but on the same host will map to the same hashkey, which means no distribution of metrics takes place. The instance is used to remedy that situation. An instance is appended to the member after the port, and separated by an equals sign, e.g. 127.0.0.1:2006=a for instance a. Instances are a concept introduced by original python carbon-cache, and need to be used in accordance with the configuration of those.

  Consistent hashes are consistent in the sense that removal of a member from the cluster should not result in a complete re-mapping of all metrics to members, but instead only add the metrics from the removed member to all remaining members, approximately each gets its fair share. The other way around, when a member is added, each member should see a subset of its metrics now being addressed to the new member. This is an important advantage over a normal hash, where each removal or addition of members (also via e.g. a change in their IP address or hostname) would cause a full re-mapping of all metrics over all available metrics.

• fnv1a_ch cluster

  The fnv1a_ch cluster is an incompatible improvement to carbon_ch introduced by carbon-c-relay. It uses a different hash technique (FNV1a) which is faster than the MD5-hashing used by carbon_ch. More importantly, fnv1a_ch clusters use both host and port to distinguish the members. This is useful when multiple targets live on the same host just separated by port.

  Since the instance property is no longer necessary with fnv1a_ch this way, it is used to completely override the host:port string that the hashkey would be calculated off. This is an important aspect, since the hashkey defines which metrics a member receives. Such override allows for many things, including masquerading old IP addresses e.g. when a machine was migrated to newer hardware. An example of this would be 10.0.0.5:2003=10.0.0.2:2003 where a machine at address 5 now receives the metrics for a machine that were at address 2.

  While using instances this way can be very useful to perform migrations in existing clusters, for newly to setup clusters, instances allow to avoid this work by using an instance from day one to detach the machine location from the metrics it receives. Consider for instance 10.0.0.1:2003=4d79d13554fa1301476c1f9fe968b0ac where a random hash as instance is used. This would allow to change port and/or ip address of the server that receives data as many times without having deal with any legacy being visible, assuming the random hash is retained. Note that since the instance name is used as full hash input, instances as a, b, etc. will likely result in poor hash distribution, as their hashes have very little input. For that reason, consider using longer and mostly differing instance names such as random hashes as used in above example for better hash distribution behaviour.

• jump_fnv1a_ch cluster

  The jump_fnv1a_ch cluster is also a consistent hash cluster like the previous two, but it does not take the member host, port or instance into account at all. This means this cluster type looks at the order in which members are defined, see also below for more on this order. Whether this is useful to you depends on your scenario. In contrast to the previous two consistent hash cluster types, the jump hash has an almost perfect balancing over the members defined in the cluster. However, this comes at the expense of not being able to remove any member but the last from the cluster without causing a complete re-mapping of all metrics over all members. What this basically means is that this hash is fine to use with constant or ever growing clusters where older nodes are never removed, but replaced instead.

  If you have a cluster where removal of old nodes takes place, the jump hash is not suitable for you. Jump hash works with servers in an ordered list. Since this order is important, it can be made explicit using the instance as used in previous cluster types. When an instance is given with the members, it will be used as sorting key. Without this instance, the order will be as given in the configuration file, which may be prone to changes when e.g. generated by some config management software. As such, it is probably a good practice to fix the order of the servers with instances such that it is explicit what the right nodes for the jump hash are. One can just use numbers for these, but be aware that sorting of 1, 2 and 10 results in 1, 10, 2, so better to use something like P0001, P0002, P0010 instead.

MATCHES

Match rules are the way to direct incoming metrics to one or more clusters. Match rules are processed top to bottom as they are defined in the file. It is possible to define multiple matches in the same rule. Each match rule can send data to one or more clusters. Since match rules "fall through" unless the stop keyword is added, carefully crafted match expression can be used to target multiple clusters or aggregations. This ability allows to replicate metrics, as well as send certain metrics to alternative clusters with careful ordering and usage of the stop keyword. The special cluster blackhole discards any metrics sent to it. This can be useful for weeding out unwanted metrics in certain cases. Because throwing metrics away is pointless if other matches would accept the same data, a match with as destination the blackhole cluster, has an implicit stop. The validation clause adds a check to the data (what comes after the metric) in the form of a regular expression. When this expression matches, the match rule will execute as if no validation clause was present. However, if it fails, the match rule is aborted, and no metrics will be sent to destinations, this is the drop behaviour. When log is used, the metric is logged to stderr. Care should be taken with the latter to avoid log flooding. When a validate clause is present, destinations need not to be present, this allows for applying a global validation rule. Note that the cleansing rules are applied before validation is done, thus the data will not have duplicate spaces. The route using clause is used to perform a temporary modification to the key used for input to the consistent hashing routines. The primary purpose is to route traffic so that appropriate data is sent to the needed aggregation instances.

REWRITES

Rewrite rules take a regular expression as input to match incoming metrics, and transform them into the desired new metric name. In the replacement, backreferences are allowed to match capture groups defined in the input regular expression. A match of server\.(x|y|z)\. allows to use e.g. role.\1. in the substitution. If needed, a notation of \g{n} can be used instead of \n where the backreference is followed by an integer, such as \g{1}100. A few caveats apply to the current implementation of rewrite rules. First, their location in the config file determines when the rewrite is performed. The rewrite is done in-place, as such a match rule before the rewrite would match the original name, a match rule after the rewrite no longer matches the original name. Care should be taken with the ordering, as multiple rewrite rules in succession can take place, e.g. a gets replaced by b and b gets replaced by c in a succeeding rewrite rule. The second caveat with the current implementation, is that the rewritten metric names are not cleansed, like newly incoming metrics are. Thus, double dots and potential dangerous characters can appear if the replacement string is crafted to produce them. It is the responsibility of the writer to make sure the metrics are clean. If this is an issue for routing, one can consider to have a rewrite-only instance that forwards all metrics to another instance that will do the routing. Obviously the second instance will cleanse the metrics as they come in. The backreference notation allows to lowercase and uppercase the replacement string with the use of the underscore (_) and carret (^) symbols following directly after the backslash. For example, role.\_1. as substitution will lowercase the contents of \1. The dot (.) can be used in a similar fashion, or followed after the underscore or caret to replace dots with underscores in the substitution. This can be handy for some situations where metrics are sent to graphite.

AGGREGATIONS

The aggregations defined take one or more input metrics expressed by one or more regular expresions, similar to the match rules. Incoming metrics are aggregated over a period of time defined by the interval in seconds. Since events may arrive a bit later in time, the expiration time in seconds defines when the aggregations should be considered final, as no new entries are allowed to be added any more. On top of an aggregation multiple aggregations can be computed. They can be of the same or different aggregation types, but should write to a unique new metric. The metric names can include back references like in rewrite expressions, allowing for powerful single aggregation rules that yield in many aggregations. When no send to clause is given, produced metrics are sent to the relay as if they were submitted from the outside, hence match and aggregation rules apply to those. Care should be taken that loops are avoided this way. For this reason, the use of the send to clause is encouraged, to direct the output traffic where possible. Like for match rules, it is possible to define multiple cluster targets. Also, like match rules, the stop keyword applies to control the flow of metrics in the matching process.

STATISTICS

The send statistics to construct is deprecated and will be removed in the next release. Use the special statistics construct instead.

The statistics construct can control a couple of things about the (internal) statistics produced by the relay. The send to target can be used to avoid router loops by sending the statistics to a certain destination cluster(s). By default the metrics are prefixed with carbon.relays.<hostname>, where hostname is determinted on startup and can be overridden using the -H argument. This prefix can be set using the prefix with clause similar to a rewrite rule target. The input match in this case is the pre-set regular expression ^(([^.]+)(\..*)?)$ on the hostname. As such, one can see that the default prefix is set by carbon.relays.\.1. Note that this uses the replace-dot-with-underscore replacement feature from rewrite rules. Given the input expression, the following match groups are available: \1 the entire hostname, \2 the short hostname and \3 the domainname (with leading dot). It may make sense to replace the default by something like carbon.relays.\_2 for certain scenarios, to always use the lowercased short hostname, which following the expression doesn't contain a dot. By default, the metrics are submitted every 60 seconds, this can be changed using the submit every <interval> seconds clause To obtain a more compatible set of values to carbon-cache.py, use the reset counters after interval clause to make values non-cumulative, that is, they will report the change compared to the previous value.

LISTENERS

The ports and protocols the relay should listen for incoming connections can be specified using the listen directive. Currently, all listeners need to be of linemode type. An optional compression or encryption wrapping can be specified for the port and optional interface given by ip address, or unix socket by file. When interface is not specified, the any interface on all available ip protocols is assumed. If no listen directive is present, the relay will use the default listeners for port 2003 on tcp and udp, plus the unix socket /tmp/.s.carbon-c-relay.2003. This typically expands to 5 listeners on an IPv6 enabled system. The default matches the behaviour of versions prior to v3.2.

INCLUDES

In case configuration becomes very long, or is managed better in separate files, the include directive can be used to read another file. The given file will be read in place and added to the router configuration at the time of inclusion. The end result is one big route configuration. Multiple include statements can be used throughout the configuration file. The positioning will influence the order of rules as normal. Beware that recursive inclusion (include from an included file) is supported, and currently no safeguards exist for an inclusion loop. For what is worth, this feature likely is best used with simple configuration files (e.g. not having include in them).

EXAMPLES

carbon-c-relay evolved over time, growing features on demand as the tool proved to be stable and fitting the job well. Below follow some annotated examples of constructs that can be used with the relay.

Clusters can be defined as much as necessary. They receive data from match rules, and their type defines which members of the cluster finally get the metric data. The simplest cluster form is a forward cluster:

cluster send-through
    forward
        10.1.0.1
    ;

Any metric sent to the send-through cluster would simply be forwarded to the server at IPv4 address 10.1.0.1. If we define multiple servers, all of those servers would get the same metric, thus:

cluster send-through
    forward
        10.1.0.1
        10.2.0.1
    ;

The above results in a duplication of metrics send to both machines. This can be useful, but most of the time it is not. The any_of cluster type is like forward, but it sends each incoming metric to any of the members. The same example with such cluster would be:

cluster send-to-any-one
    any_of 10.1.0.1:2010 10.1.0.1:2011;

This would implement a multipath scenario, where two servers are used, the load between them is spread, but should any of them fail, all metrics are sent to the remaining one. This typically works well for upstream relays, or for balancing carbon-cache processes running on the same machine. Should any member become unavailable, for instance due to a rolling restart, the other members receive the traffic. If it is necessary to have true fail-over, where the secondary server is only used if the first is down, the following would implement that:

cluster try-first-then-second
    failover 10.1.0.1:2010 10.1.0.1:2011;

These types are different from the two consistent hash cluster types:

cluster graphite
    carbon_ch
        127.0.0.1:2006=a
        127.0.0.1:2007=b
        127.0.0.1:2008=c
    ;

If a member in this example fails, all metrics that would go to that member are kept in the queue, waiting for the member to return. This is useful for clusters of carbon-cache machines where it is desirable that the same metric ends up on the same server always. The carbon_ch cluster type is compatible with carbon-relay consistent hash, and can be used for existing clusters populated by carbon-relay. For new clusters, however, it is better to use the fnv1a_ch cluster type, for it is faster, and allows to balance over the same address but different ports without an instance number, in constrast to carbon_ch.

Because we can use multiple clusters, we can also replicate without the use of the forward cluster type, in a more intelligent way:

cluster dc-old
    carbon_ch replication 2
        10.1.0.1
        10.1.0.2
        10.1.0.3
    ;
cluster dc-new1
    fnv1a_ch replication 2
        10.2.0.1
        10.2.0.2
        10.2.0.3
    ;
cluster dc-new2
    fnv1a_ch replication 2
        10.3.0.1
        10.3.0.2
        10.3.0.3
    ;

match *
    send to dc-old
    ;
match *
    send to
        dc-new1
        dc-new2
    stop
    ;

In this example all incoming metrics are first sent to dc-old, then dc-new1 and finally to dc-new2. Note that the cluster type of dc-old is different. Each incoming metric will be send to 2 members of all three clusters, thus replicating to in total 6 destinations. For each cluster the destination members are computed independently. Failure of clusters or members does not affect the others, since all have individual queues. The above example could also be written using three match rules for each dc, or one match rule for all three dcs. The difference is mainly in performance, the number of times the incoming metric has to be matched against an expression. The stop rule in dc-new match rule is not strictly necessary in this example, because there are no more following match rules. However, if the match would target a specific subset, e.g. ^sys\., and more clusters would be defined, this could be necessary, as for instance in the following abbreviated example:

cluster dc1-sys ... ;
cluster dc2-sys ... ;

cluster dc1-misc ... ;
cluster dc2-misc ... ;

match ^sys\. send to dc1-sys;
match ^sys\. send to dc2-sys stop;

match * send to dc1-misc;
match * send to dc2-misc stop;

As can be seen, without the stop in dc2-sys' match rule, all metrics starting with sys. would also be send to dc1-misc and dc2-misc. It can be that this is desired, of course, but in this example there is a dedicated cluster for the sys metrics.

Suppose there would be some unwanted metric that unfortunately is generated, let's assume some bad/old software. We don't want to store this metric. The blackhole cluster is suitable for that, when it is harder to actually whitelist all wanted metrics. Consider the following:

match
        some_legacy1$
        some_legacy2$
    send to blackhole
    stop;

This would throw away all metrics that end with some_legacy, that would otherwise be hard to filter out. Since the order matters, it can be used in a construct like this:

cluster old ... ;
cluster new ... ;

match * send to old;

match unwanted send to blackhole stop;

match * send to new;

In this example the old cluster would receive the metric that's unwanted for the new cluster. So, the order in which the rules occur does matter for the execution.

Validation can be used to ensure the data for metrics is as expected. A global validation for just number (no floating point) values could be:

match *
    validate ^[0-9]+\ [0-9]+$ else drop
    ;

(Note the escape with backslash \ of the space, you might be able to use \s or [:space:] instead, this depends on your configured regex implementation.)

The validation clause can exist on every match rule, so in principle, the following is valid:

match ^foo
    validate ^[0-9]+\ [0-9]+$ else drop
    send to integer-cluster
    ;
match ^foo
    validate ^[0-9.e+-]+\ [0-9.e+-]+$ else drop
    send to float-cluster
    stop;

Note that the behaviour is different in the previous two examples. When no send to clusters are specified, a validation error makes the match behave like the stop keyword is present. Likewise, when validation passes, processing continues with the next rule. When destination clusters are present, the match respects the stop keyword as normal. When specified, processing will always stop when specified so. However, if validation fails, the rule does not send anything to the destination clusters, the metric will be dropped or logged, but never sent.

The relay is capable of rewriting incoming metrics on the fly. This process is done based on regular expressions with capture groups that allow to substitute parts in a replacement string. Rewrite rules allow to cleanup metrics from applications, or provide a migration path. In it's simplest form a rewrite rule looks like this:

rewrite ^server\.(.+)\.(.+)\.([a-zA-Z]+)([0-9]+)
    into server.\_1.\2.\3.\3\4
    ;

In this example a metric like server.DC.role.name123 would be transformed into server.dc.role.name.name123. For rewrite rules hold the same as for matches, that their order matters. Hence to build on top of the old/new cluster example done earlier, the following would store the original metric name in the old cluster, and the new metric name in the new cluster:

rewrite ^server\.(.+)\.(.+)\.([a-zA-Z]+)([0-9]+)
    into server.\_1.\2.\3.\3\4
    ;
rewrite ^server\.(.+)\.(.+)\.([a-zA-Z]+)([0-9]+)
    into server.\g{_1}.\g{2}.\g{3}.\g{3}\g{4}
    ;

The alternate syntax for backreference notation using g\{n} instead of \n notation shown above. Both rewrite rules are identical.

match * send to old;

rewrite ... ;

match * send to new;

Note that after the rewrite, the original metric name is no longer available, as the rewrite happens in-place.

Aggregations are probably the most complex part of carbon-c-relay. Two ways of specifying aggregates are supported by carbon-c-relay. The first, static rules, are handled by an optimiser which tries to fold thousands of rules into groups to make the matching more efficient. The second, dynamic rules, are very powerful compact definitions with possibly thousands of internal instantiations. A typical static aggregation looks like:

aggregate
        ^sys\.dc1\.somehost-[0-9]+\.somecluster\.mysql\.replication_delay
        ^sys\.dc2\.somehost-[0-9]+\.somecluster\.mysql\.replication_delay
    every 10 seconds
    expire after 35 seconds
    timestamp at end of bucket
    compute sum write to
        mysql.somecluster.total_replication_delay
    compute average write to
        mysql.somecluster.average_replication_delay
    compute max write to
        mysql.somecluster.max_replication_delay
    compute count write to
        mysql.somecluster.replication_delay_metric_count
    ;

In this example, four aggregations are produced from the incoming matching metrics. In this example we could have written the two matches as one, but for demonstration purposes we did not. Obviously they can refer to different metrics, if that makes sense. The every 10 seconds clause specifies in what interval the aggregator can expect new metrics to arrive. This interval is used to produce the aggregations, thus each 10 seconds 4 new metrics are generated from the data received sofar. Because data may be in transit for some reason, or generation stalled, the expire after clause specifies how long the data should be kept before considering a data bucket (which is aggregated) to be complete. In the example, 35 was used, which means after 35 seconds the first aggregates are produced. It also means that metrics can arrive 35 seconds late, and still be taken into account. The exact time at which the aggregate metrics are produced is random between 0 and interval (10 in this case) seconds after the expiry time. This is done to prevent thundering herds of metrics for large aggregation sets. The timestamp that is used for the aggregations can be specified to be the start, middle or end of the bucket. Original carbon-aggregator.py uses start, while carbon-c-relay's default has always been end. The compute clauses demonstrate a single aggregation rule can produce multiple aggregates, as often is the case. Internally, this comes for free, since all possible aggregates are always calculated, whether or not they are used. The produced new metrics are resubmitted to the relay, hence matches defined before in the configuration can match output of the aggregator. It is important to avoid loops, that can be generated this way. In general, splitting aggregations to their own carbon-c-relay instance, such that it is easy to forward the produced metrics to another relay instance is a good practice.

The previous example could also be written as follows to be dynamic:

aggregate
        ^sys\.dc[0-9].(somehost-[0-9]+)\.([^.]+)\.mysql\.replication_delay
    every 10 seconds
    expire after 35 seconds
    compute sum write to
        mysql.host.\1.replication_delay
    compute sum write to
        mysql.host.all.replication_delay
    compute sum write to
        mysql.cluster.\2.replication_delay
    compute sum write to
        mysql.cluster.all.replication_delay
    ;

Here a single match, results in four aggregations, each of a different scope. In this example aggregation based on hostname and cluster are being made, as well as the more general all targets, which in this example have both identical values. Note that with this single aggregation rule, both per-cluster, per-host and total aggregations are produced. Obviously, the input metrics define which hosts and clusters are produced.

With use of the send to clause, aggregations can be made more intuitive and less error-prone. Consider the below example:

cluster graphite fnv1a_ch ip1 ip2 ip3;

aggregate ^sys\.somemetric
    every 60 seconds
    expire after 75 seconds
    compute sum write to
        sys.somemetric
    send to graphite
    stop
    ;

match * send to graphite;

It sends all incoming metrics to the graphite cluster, except the sys.somemetric ones, which it replaces with a sum of all the incoming ones. Without a stop in the aggregate, this causes a loop, and without the send to, the metric name can't be kept its original name, for the output now directly goes to the cluster.

When configuring cluster you might want to check how the metrics will be routed and hashed. That's what the -t flag is for. For the following configuration:

cluster graphite_swarm_odd
    fnv1a_ch replication 1
        host01.dom:2003=31F7A65E315586AC198BD798B6629CE4903D089947
        host03.dom:2003=9124E29E0C92EB63B3834C1403BD2632AA7508B740
        host05.dom:2003=B653412CD96B13C797658D2C48D952AEC3EB667313
;

cluster graphite_swarm_even
    fnv1a_ch replication 1
        host02.dom:2003=31F7A65E315586AC198BD798B6629CE4903D089947
        host04.dom:2003=9124E29E0C92EB63B3834C1403BD2632AA7508B740
        host06.dom:2003=B653412CD96B13C797658D2C48D952AEC3EB667313

;

match *
    send to
        graphite_swarm_odd
        graphite_swarm_even
    stop
;

Running the command: echo "my.super.metric" | carbon-c-relay -f config.conf -t, will result in:

[...]
match
    * -> my.super.metric
    fnv1a_ch(graphite_swarm_odd)
        host03.dom:2003
    fnv1a_ch(graphite_swarm_even)
        host04.dom:2003
    stop

You now know that your metric my.super.metric will be hashed and arrive on the host03 and host04 machines. Adding the -d flag will increase the amount of information by showing you the hashring

STATISTICS

When carbon-c-relay is run without -d or -s arguments, statistics will be produced. By default they are sent to the relay itself in the form of carbon.relays.<hostname>.*. See the statistics construct to override this prefix, sending interval and values produced. While many metrics have a similar name to what carbon-cache.py would produce, their values are likely different. By default, most values are running counters which only increase over time. The use of the nonNegativeDerivative() function from graphite is useful with these.

The following metrics are produced under the carbon.relays.<hostname> namespace:

• metricsReceived

  The number of metrics that were received by the relay. Received here means that they were seen and processed by any of the dispatchers.

• metricsSent

  The number of metrics that were sent from the relay. This is a total count for all servers combined. When incoming metrics are duplicated by the cluster configuration, this counter will include all those duplications. In other words, the amount of metrics that were successfully sent to other systems. Note that metrics that are processed (received) but still in the sending queue (queued) are not included in this counter.

• metricsDiscarded

  The number of input lines that were not considered to be a valid metric. Such lines can be empty, only containing whitespace, or hitting the limits given for max input length and/or max metric length (see -m and -M options).

• metricsQueued

  The total number of metrics that are currently in the queues for all the server targets. This metric is not cumulative, for it is a sample of the queue size, which can (and should) go up and down. Therefore you should not use the derivative function for this metric.

• metricsDropped

  The total number of metric that had to be dropped due to server queues overflowing. A queue typically overflows when the server it tries to send its metrics to is not reachable, or too slow in ingesting the amount of metrics queued. This can be network or resource related, and also greatly depends on the rate of metrics being sent to the particular server.

• metricsBlackholed

  The number of metrics that did not match any rule, or matched a rule with blackhole as target. Depending on your configuration, a high value might be an indication of a misconfiguration somewhere. These metrics were received by the relay, but never sent anywhere, thus they disappeared.

• metricStalls

  The number of times the relay had to stall a client to indicate that the downstream server cannot handle the stream of metrics. A stall is only performed when the queue is full and the server is actually receptive of metrics, but just too slow at the moment. Stalls typically happen during micro-bursts, where the client typically is unaware that it should stop sending more data, while it is able to.

• connections

  The number of connect requests handled. This is an ever increasing number just counting how many connections were accepted.

• disconnects

  The number of disconnected clients. A disconnect either happens because the client goes away, or due to an idle timeout in the relay. The difference between this metric and connections is the amount of connections actively held by the relay. In normal situations this amount remains within reasonable bounds. Many connections, but few disconnections typically indicate a possible connection leak in the client. The idle connections disconnect in the relay here is to guard against resource drain in such scenarios.

• dispatch_wallTime_us

  The number of microseconds spent by the dispatchers to do their work. In particular on multi-core systems, this value can be confusing, however, it indicates how long the dispatchers were doing work handling clients. It includes everything they do, from reading data from a socket, cleaning up the input metric, to adding the metric to the appropriate queues. The larger the configuration, and more complex in terms of matches, the more time the dispatchers will spend on the cpu. But also time they do /not/ spend on the cpu is included in this number. It is the pure wallclock time the dispatcher was serving a client.

• dispatch_sleepTime_us

  The number of microseconds spent by the dispatchers sleeping waiting for work. When this value gets small (or even zero) the dispatcher has so much work that it doesn't sleep any more, and likely can't process the work in a timely fashion any more. This value plus the wallTime from above sort of sums up to the total uptime taken by this dispatcher. Therefore, expressing the wallTime as percentage of this sum gives the busyness percentage draining all the way up to 100% if sleepTime goes to 0.

• server_wallTime_us

  The number of microseconds spent by the servers to send the metrics from their queues. This value includes connection creation, reading from the queue, and sending metrics over the network.

• dispatcherX

  For each indivual dispatcher, the metrics received and blackholed plus the wall clock time. The values are as described above.

• destinations.X

  For all known destinations, the number of dropped, queued and sent metrics plus the wall clock time spent. The values are as described above.

• aggregators.metricsReceived

  The number of metrics that were matched an aggregator rule and were accepted by the aggregator. When a metric matches multiple aggregators, this value will reflect that. A metric is not counted when it is considered syntactically invalid, e.g. no value was found.

• aggregators.metricsDropped

  The number of metrics that were sent to an aggregator, but did not fit timewise. This is either because the metric was too far in the past or future. The expire after clause in aggregate statements controls how long in the past metric values are accepted.

• aggregators.metricsSent

  The number of metrics that were sent from the aggregators. These metrics were produced and are the actual results of aggregations.

BUGS

Please report them at: https://github.com/grobian/carbon-c-relay/issues

AUTHOR

Fabian Groffen <grobian@gentoo.org>

SEE ALSO

All other utilities from the graphite stack.

This project aims to be a fast replacement of the original Carbon relay. carbon-c-relay aims to deliver performance and configurability. Carbon is single threaded, and sending metrics to multiple consistent-hash clusters requires chaining of relays. This project provides a multithreaded relay which can address multiple targets and clusters for each and every metric based on pattern matches.

There are a couple more replacement projects out there, which are carbon-relay-ng and graphite-relay.

Compared to carbon-relay-ng, this project does provide carbon's consistent-hash routing. graphite-relay, which does this, however doesn't do metric-based matches to direct the traffic, which this project does as well. To date, carbon-c-relay can do aggregations, failover targets and more.

ACKNOWLEDGEMENTS

This program was originally developed for Booking.com, which approved that the code was published and released as Open Source on GitHub, for which the author would like to express his gratitude. Development has continued since with the help of many contributors suggesting features, reporting bugs, adding patches and more to make carbon-c-relay into what it is today.