HDFS-15382. Split one FsDatasetImpl lock to block pool grain locks. #3941
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
💔 -1 overall
This message was automatically generated. |
MingXiangLi
force-pushed
the
trunk-block-pool-lock
branch
from
c37adaf
to
1cfe96b
Compare
|
💔 -1 overall
This message was automatically generated. |
|
@sodonnel sorry to interrupt, this patch seems very promising. Any comment about this patch? |
Hexiaoqiao
reviewed
Feb 21, 2022
There was a problem hiding this comment.
Thanks @MingXiangLi for your works. Have a quick glance and leave some nit comments inline, I think most of the changes look good to me. Consider it is big changes here we should involve more guys to review and evaluate this improvement.
BTW, the check annotations warning seem not related to this changes. I will trigger the Jenkins again, and let's wait what jenkins says.
| @@ -504,15 +503,13 @@ private void createWorkPlan(NodePlan plan) throws DiskBalancerException { | |||
| Map<String, String> storageIDToVolBasePathMap = new HashMap<>(); | |||
| FsDatasetSpi.FsVolumeReferences references; | |||
| try { | |||
| try(AutoCloseableLock lock = this.dataset.acquireDatasetReadLock()) { | |||
There was a problem hiding this comment.
Just suggest to keep it and improve to #dataSetLock as other improvement. We should create another jira to remove it if this is not necessary now.
There was a problem hiding this comment.
this method just get volumes snapshot so no need to acquire dataset lock.
| @@ -464,42 +427,40 @@ public AutoCloseableLock acquireDatasetReadLock() { | |||
| * Activate a volume to serve requests. | |||
| * @throws IOException if the storage UUID already exists. | |||
| */ | |||
| private void activateVolume( | |||
| private synchronized void activateVolume( | |||
There was a problem hiding this comment.
Just suggest to add annotation to describe that we use #synchronized to protect volume instance and use datasetLock to protect blockpool instance.
| @@ -3654,18 +3661,21 @@ public int getVolumeCount() { | |||
| } | |||
|
|
|||
| void stopAllDataxceiverThreads(FsVolumeImpl volume) { | |||
| try (AutoCloseableLock lock = datasetWriteLock.acquire()) { | |||
| @@ -232,118 +232,6 @@ public void setUp() throws IOException { | |||
| assertEquals(0, dataset.getNumFailedVolumes()); | |||
| } | |||
|
|
|||
| @Test(timeout=10000) | |||
There was a problem hiding this comment.
I do not get why delete this unit test here.
There was a problem hiding this comment.
this old lock model no longer useful
|
💔 -1 overall
This message was automatically generated. |
jojochuang
reviewed
Feb 22, 2022
| @@ -3230,7 +3235,9 @@ void transferReplicaForPipelineRecovery(final ExtendedBlock b, | |||
| final BlockConstructionStage stage; | |||
|
|
|||
| //get replica information | |||
| try(AutoCloseableLock lock = data.acquireDatasetReadLock()) { | |||
|
|
|||
| try(AutoCloseableLock lock = dataSetLockManager.writeLock( | |||
There was a problem hiding this comment.
this was a read lock before. Any idea why it is made a write lock now?
There was a problem hiding this comment.
this should be read lock.I write wrong here.
| } | ||
| references.close(); |
There was a problem hiding this comment.
It would be better to instantiate the references object in a try .. with block to ensure it is closed properly even with an exception,
There was a problem hiding this comment.
ok,I will fix this later
| @@ -602,7 +563,7 @@ public void removeVolumes( | |||
| new ArrayList<>(storageLocsToRemove); | |||
| Map<String, List<ReplicaInfo>> blkToInvalidate = new HashMap<>(); | |||
| List<String> storageToRemove = new ArrayList<>(); | |||
| try (AutoCloseableLock lock = datasetWriteLock.acquire()) { | |||
| synchronized (this) { | |||
There was a problem hiding this comment.
how about making the entire method synchronized instead?
There was a problem hiding this comment.
This object lock is protect data structure related volumes like add and remove.So just replace dataset global write lock.HDFS-9445 fix problem if use synchronized to whole method.
| @@ -117,7 +121,7 @@ ReplicaInfo get(String bpid, long blockId) { | |||
| ReplicaInfo add(String bpid, ReplicaInfo replicaInfo) { | |||
| checkBlockPool(bpid); | |||
| checkBlock(replicaInfo); | |||
| try (AutoCloseableLock l = writeLock.acquire()) { | |||
| try (AutoCloseDataSetLock l = lockManager.readLock(LockLevel.BLOCK_POOl, bpid)) { | |||
There was a problem hiding this comment.
why is read lock used here?
LightWeightResizableGSet is not thread-safe.
There was a problem hiding this comment.
in new version I have change LightWeightResizableGSet to make it thread-safe.
tomscut
reviewed
Feb 22, 2022
| @@ -1850,7 +1828,8 @@ public ReplicaHandler createTemporary(StorageType storageType, | |||
| ReplicaInfo lastFoundReplicaInfo = null; | |||
| boolean isInPipeline = false; | |||
| do { | |||
| try (AutoCloseableLock lock = datasetWriteLock.acquire()) { | |||
| try (AutoCloseableLock lock = lockManager.readLock(LockLevel.BLOCK_POOl, | |||
There was a problem hiding this comment.
Here write lock is changed to read lock, is there any problem?
| @@ -184,7 +203,7 @@ void mergeAll(ReplicaMap other) { | |||
| ReplicaInfo remove(String bpid, Block block) { | |||
| checkBlockPool(bpid); | |||
| checkBlock(block); | |||
| try (AutoCloseableLock l = writeLock.acquire()) { | |||
| try (AutoCloseDataSetLock l = lockManager.readLock(LockLevel.BLOCK_POOl, bpid)) { | |||
There was a problem hiding this comment.
I agree with @Hexiaoqiao . We should keep the read/write lock. Once block Pool lock is introduced, we can discuss whether write locks need to be changed to read locks in other JIRA.
There was a problem hiding this comment.
Now the LightWeightResizableGSet is thread-safe, so most operate in ReplicaMap just need block pool read lock.
|
@jojochuang @tomscut The previous PR has merge refer to #3900. At this PR it improve LightWeightResizableGSet to thread-safe and create LockManager mode. Please reference. |
jojochuang
reviewed
Feb 23, 2022
| * | ||
| * http://www.apache.org/licenses/LICENSE-2.0 | ||
| * | ||
| * <p> |
There was a problem hiding this comment.
Please leave the license boiler plate text unchanged.
Thanks @Hexiaoqiao for your reply. I'll look at this PR later. |
MingXiangLi
force-pushed
the
trunk-block-pool-lock
branch
from
1cfe96b
to
cda5cbc
Compare
|
renew the pr improvement TODO |
|
🎊 +1 overall
This message was automatically generated. |
MingXiangLi
force-pushed
the
trunk-block-pool-lock
branch
from
cda5cbc
to
3586f00
Compare
|
🎊 +1 overall
This message was automatically generated. |
Hexiaoqiao
approved these changes
Mar 1, 2022
There was a problem hiding this comment.
Thanks @MingXiangLi , The latest version makes sense to me. +1 from my side.
cc @jojochuang @tomscut Would you mind to give another reviews. Thanks.
|
Thanks @MingXiangLi and @Hexiaoqiao . This is a great idea to introduce fine-grained locking for datanode. In fact, we introduced original PR of HDFS-15380 and The latest version looks good to me. |
tomscut
approved these changes
Mar 1, 2022
|
Thanks @tomscut for your reviews. Will commit to trunk wait for two days if no more other comments here. Thanks. |
asfgit
pushed a commit
that referenced
this pull request
Mar 12, 2022
…3941). Contributed by limingxiang. Signed-off-by: He Xiaoqiao <hexiaoqiao@apache.org> Signed-off-by: litao <tomleescut@gmail.com>
|
Committed to trunk. Thanks @MingXiangLi for your contributions! Thanks @tomscut and @jojochuang for your reviews! |
HarshitGupta11
pushed a commit
to HarshitGupta11/hadoop
that referenced
this pull request
Nov 28, 2022
…pache#3941). Contributed by limingxiang. Signed-off-by: He Xiaoqiao <hexiaoqiao@apache.org> Signed-off-by: litao <tomleescut@gmail.com>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Design doc
https://drive.google.com/file/d/1eaE8vSEhIli0H3j2eDiPJNYuKAC0MFgu/view?usp=sharing
https://issues.apache.org/jira/browse/HDFS-15382