SNMP tables are not collected #10445
Comments
|
@Takuto88 2 suggestions:
|
|
Thanks for the suggestions
I've tried that but unfortunately, that does not seem to do anything
That's what we'll do for now. Just wanted to report this so that the devs know that something might have broken here. |
|
+1 |
|
I the YATE-MIB dependent on any of the MIBs that could not be loaded? |
|
YATE-MIB relies only on NULL-TEAM-MIB and the usual SNMPv2-SMI and such so should be fine re. dependencies. I don't actually see any errors in the logs as we have seen in other cases (i.e. #10444 where there is a translation error). I wonder if there is something else going on here.. @Takuto88 are you in a position to get a tcpdump of the SNMP traffic from telegraf running 1.21.2 to your YATE server - with full packet size? (i.e. set the Was the log provided generated from the exact config you included (modulo credentials, of course)? |
Yes I am in a position to provide the trace. Please find it attached here: trace.pcap.zip The port 1161 is correct. When looking at the trace in Wireshark, one must remember to change the SNMP UDP Port via Preferences -> Protocols -> SNMP. Otherwise Wireshark will not decode the SNMP data. As far as I can tell, the Tables are collected via SNMP (e.g. get-bulk request for 1.3.6.1.4.1.34501.1.6.2.2.1.2). But there is nothing in the debug log and no data is pushed to Influxdb which is really odd.
Yes that's a copy-paste of the relevant part. We do collect more stuff that I've omitted to keep the example concise. I can provide the full configuration (minus secrets like passwords 😉 ) if you feel that it may be relevant. The same config works with 1.20.4 just fine (except for the new |
I've found my problem (and maybe the same for others ?). I suppose the gosmi lib needs a "good" SNMPv2-SMI file to do the job...
|
|
@Takuto88, maybe you have the same problem I had with a bad version of SNMPv2-SMI.txt. Of course I don't have a device that responds to these YATE oids... but I'm pretty sure (I hope) it'll work for you too. |
|
@igit Thank you for your investigation. Unfortunately, I don't see a different behaviour than before when loading just these files. Telegraf still scrapes these tables but does not push their contents to influx for whatever reason. |
telegraf-tiger
bot
removed
the
waiting for response
|
Could you also give the results of running telegraf with |
|
@Hipska Sure thing. Here you go: |
|
Okay, indeed nothing in there. (might have been an issue related to your filtering on InfluxDB) But I also seem some output that was not configured, so could you please provide the exact
So you confirm the table gets collected when specifying one or more fields individually? @MyaLongmire could you have a look on this? Seems like table init has some issues in this case.. |
telegraf-tiger
bot
removed
the
waiting for response
|
Hello! I am closing this issue due to inactivity. I hope you were able to resolve your problem, if not please try posting this question in our Community Slack or Community Page. Thank you! |
telegraf-tiger
bot
removed
the
waiting for response
|
Hello! I am closing this issue due to inactivity. I hope you were able to resolve your problem, if not please try posting this question in our Community Slack or Community Page. Thank you! |
telegraf-tiger
bot
removed
the
waiting for response
|
Sorry about the noise from the tiger bot, this is a bug I'll look into why it wants to close this issue. |
|
Hi @Takuto88 and @igit, we recently merged a change that adds back MIB translation by netsnmp (snmptranslate and snmptable). See #10802 Whether the specific problem you're hitting is caused by and old/incomplete SNMPv2-SMI file or whatever, if you had a working configuration with 1.20.4 it should work again with recent nightly builds. Could you test with a nightly build to see if this change fixes your problem and lets you upgrade? We'd like to release it with 1.23.0 which is planned for Wednesday, March 23. If you could report what you find it would help us flush out bugs and gain confidence in the change. |
|
The problem here seems to be with non working table init. Not directly related to MIB loading. |
|
@Hipska Table init and mib loading were both changed when we switched to gosmi. 10802 switched both back to the 1.20.4 code. Since downgrading to 1.20.4 worked in January for Takuto88, it seems likely to me that a build with 10802 will also work. It's definitely worth retesting. If they test and find the nightly build doesn't work for them, that's a data point that's very valuable for debugging. |
|
@Takuto88 Could you please come back and let us know if this is resolved by using the latest builds of Telegraf? |
@reimda , for me the issue was a problem of an old MIB file not recognized by gosmi parser. When I updated my files and test the parsing, all was ok. I tested the file used by @Takuto88, for me the gosmi parsing was ok. Not sure the problem was here... But, I looked at the PR #10802, I think it's a good idea to have the two parsers (gosmi and Net-SNMP) in the code, and let's choose the user (with conf translator = "gosmi"). I'll check it in next version ! |
telegraf-tiger
bot
removed
the
waiting for response
|
Just a quick update from me as the issue reporter: We could upgrade from Telegraf v1.20.4 to v1.22.0 without any changes to our config and it just works 🎉 . Thanks to everyone involved for fixing this 💐 . P.S.: I am sorry for the delayed response. It took us a while to get back to this and verify the fix. |
|
@Takuto88, thanks for coming back with an update! Happy to hear your issue is fixed 🕺 |
Relevent telegraf.conf
Logs from Telegraf
System info
Telegraf 1.21.2 - Debian 10 (Buster)
Docker
No response
Steps to reproduce
Expected behavior
The collected metrics should be pushed to influxdb
Actual behavior
The metrics are collected via SNMP (we've seen this by watching
tcpdump) but telegraf never pushes them to influxdb.Additional info
Downgrading to 1.20.4 resolved the issue for us. We suspect that the snmptranslate to gosnmp migration breaks something.
We're using custom MIBS for a Software called YATE. These are configued in telegraf here:
The contents of the MIBS folder
/usr/local/share/yate/dataare as follows:NULL-TEAM-MIB.txt
YATE-MIB.txt
snmp_mib.conf
Note: The file
snmp_mib.confseems to fail to load according to the logfile:2022-01-14T09:39:37Z W! [inputs.snmp] module snmp_mib.conf could not be loaded.Using
YATE-MIB::linksetTableas oid or the numeric path.1.3.6.1.4.1.34501.1.6.2.2makes no differenceThe SNMP Table in question looks like this:
Other things we observed: If we configure individual values from tables, things seem to work fine. But tables as a whole can lo longer be collected.
The text was updated successfully, but these errors were encountered: