[fix][broker] timeout when broker registry hangs and monitor broker registry (ExtensibleLoadManagerImpl only)

[heesung-sn]

Motivation

• metadata store could fail the broker lock registration. In this case, we should retry and also monitor the registry periodically and try to register it if not registered yet.
• also we need to verify if the broker is healthy or not before starting the orphan bundle cleanups, as the broker might receive "out-dated" notifications.

Modifications

• add monitor logic to periodically check if the broker is registered and to try to register it.
• add timeout logic when the broker re-registration hangs.
• add retry logic when the broker re-registration fails.
• add health check logic before starting the orphan bundle cleanup jobs and canceling the cleanup jobs.
• add brokerId in broker health admin api

Verifying this change

• Make sure that the change passes the CI checks.

Does this pull request potentially affect one of the following parts:

If the box was checked, please highlight the changes

• Dependencies (add or upgrade a dependency)
• The public API
• The schema
• The default values of configurations
• The threading model
• The binary protocol
• The REST endpoints
• The admin CLI options
• The metrics
• Anything that affects deployment

Documentation

• doc
• doc-required
• doc-not-needed
• doc-complete

Matching PR in forked repository

PR in forked repository: heesung-sn#82

[lhotari]

@heesung-sn please add the release labels that this applies to.

[heesung-sn]

I realized we need to add more validation upon this broker delete/create notifications as these notification could be "out-dated".

[lhotari]

@heesung-sn I ran into a hanging test recently. Is this something that is related?

ZkSessionExpireTest.cleanup hangs in ModularLoadManagerImpl.disableBroker

"main" #1 [2383] prio=5 os_prio=0 cpu=101566.56ms elapsed=3520.41s tid=0x00007f3618030a30 nid=2383 waiting on condition  [0x00007f3621f54000]
   java.lang.Thread.State: WAITING (parking)
	at jdk.internal.misc.Unsafe.park(java.base@21.0.4/Native Method)
	- parking to wait for  <0x0000100041604dc8> (a java.util.concurrent.CompletableFuture$Signaller)
	at java.util.concurrent.locks.LockSupport.park(java.base@21.0.4/LockSupport.java:221)
	at java.util.concurrent.CompletableFuture$Signaller.block(java.base@21.0.4/CompletableFuture.java:1864)
	at java.util.concurrent.ForkJoinPool.unmanagedBlock(java.base@21.0.4/ForkJoinPool.java:3780)
	at java.util.concurrent.ForkJoinPool.managedBlock(java.base@21.0.4/ForkJoinPool.java:3725)
	at java.util.concurrent.CompletableFuture.waitingGet(java.base@21.0.4/CompletableFuture.java:1898)
	at java.util.concurrent.CompletableFuture.join(java.base@21.0.4/CompletableFuture.java:2117)
	at org.apache.pulsar.broker.loadbalance.impl.ModularLoadManagerImpl.disableBroker(ModularLoadManagerImpl.java:603)
	at org.apache.pulsar.broker.loadbalance.impl.ModularLoadManagerWrapper.disableBroker(ModularLoadManagerWrapper.java:47)
	at org.apache.pulsar.broker.service.BrokerService.unloadNamespaceBundlesGracefully(BrokerService.java:949)
	at org.apache.pulsar.broker.service.BrokerService.unloadNamespaceBundlesGracefully(BrokerService.java:936)
	at org.apache.pulsar.broker.PulsarService.closeAsync(PulsarService.java:520)
	at org.apache.pulsar.broker.PulsarService.close(PulsarService.java:479)
	at org.apache.pulsar.broker.service.NetworkErrorTestBase.cleanup(NetworkErrorTestBase.java:215)
	at org.apache.pulsar.broker.service.ZkSessionExpireTest.cleanup(ZkSessionExpireTest.java:50)

Thread dump: https://gist.github.com/lhotari/68578ba1354d4833af489b9ac3658c8c
Analysed in jstack.review: https://jstack.review/?https://gist.github.com/lhotari/68578ba1354d4833af489b9ac3658c8c#tda_1_dump

I created #23388 and #23389 for ZkSessionExpireTest flakiness issues that appeared in subsequent builds.

[heesung-sn]

@heesung-sn I ran into a hanging test recently. Is this something that is related?

Potentially, I think this could be related, regarding why the broker registry lock creation didn't go through. Unfortunately, we couldn't confirm this because we haven't put error log when creating the lock.

I see a potential deadlock report in NativeReferenceQueue$Lock. I wonder if java 21 has any bugs there.

[codecov-commenter]

Codecov Report

Attention: Patch coverage is 89.89899% with 10 lines in your changes missing coverage. Please review.

Project coverage is 74.54%. Comparing base (bbc6224) to head (051295d).
Report is 625 commits behind head on master.

Files with missing lines	Patch %	Lines
...xtensions/channel/ServiceUnitStateChannelImpl.java	84.78%	5 Missing and 2 partials ⚠️
...ker/loadbalance/extensions/BrokerRegistryImpl.java	94.11%	1 Missing and 1 partial ⚠️
...ache/pulsar/client/admin/internal/BrokersImpl.java	85.71%	1 Missing ⚠️

Additional details and impacted files

@@             Coverage Diff              @@
##             master   #23382      +/-   ##
============================================
+ Coverage     73.57%   74.54%   +0.97%     
- Complexity    32624    34015    +1391     
============================================
  Files          1877     1936      +59     
  Lines        139502   145365    +5863     
  Branches      15299    15893     +594     
============================================
+ Hits         102638   108366    +5728     
+ Misses        28908    28681     -227     
- Partials       7956     8318     +362     

Flag	Coverage Δ
inttests	27.49% <24.24%> (+2.91%)	⬆️
systests	24.48% <0.00%> (+0.16%)	⬆️
unittests	73.88% <89.89%> (+1.04%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
...g/apache/pulsar/broker/admin/impl/BrokersBase.java	73.02% <100.00%> (+0.57%)	⬆️
...dbalance/extensions/ExtensibleLoadManagerImpl.java	79.25% <100.00%> (-0.83%)	⬇️
...ache/pulsar/client/admin/internal/BrokersImpl.java	88.23% <85.71%> (+0.73%)	⬆️
...ker/loadbalance/extensions/BrokerRegistryImpl.java	84.21% <94.11%> (-0.56%)	⬇️
...xtensions/channel/ServiceUnitStateChannelImpl.java	85.35% <84.78%> (+0.04%)	⬆️

... and 601 files with indirect coverage changes

Comments resolved

[lhotari]

@heesung-sn Cherry-picking this to branch-3.0 doesn't get applied cleanly. Would you be able to handle cherry-picking the required changes to branch-3.0?