[
https://issues.apache.org/jira/browse/CASSANDRA-19776?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17952492#comment-17952492
]
Cameron Zemek commented on CASSANDRA-19776:
-------------------------------------------
`Since the issue is currently limited to the metrics` – is it though?
I have given up trying to come up with a test that reproduces the issue
reliably. From my understanding it's a race condition where the sstable is
obsolete but is still part of CANONICAL. There are numerous calls to
selectAndReference on the CANONICAL set that potentially could have the same
spin lock issue. But considering I have not been able to come up with test and
have not seen this spin lock occur outside of the metric collection when
compactions occur perhaps it is isolated. However, if it is isolated to this
metric, why is that the case?
Why do only these metrics grab references, and all the other table metrics
don't? What happens when collecting metrics on sstables that are no longer
valid (since they can be removed when no references remain)?
> Spinning trying to capture readers
> ----------------------------------
>
> Key: CASSANDRA-19776
> URL: https://issues.apache.org/jira/browse/CASSANDRA-19776
> Project: Apache Cassandra
> Issue Type: Bug
> Components: Legacy/Core
> Reporter: Cameron Zemek
> Assignee: Stefan Miklosovic
> Priority: Normal
> Fix For: 4.0.x, 4.1.x, 5.0.x, 5.x
>
> Attachments: extract.log
>
> Time Spent: 20m
> Remaining Estimate: 0h
>
> On a handful of clusters we are noticing Spin locks occurring. I traced back
> all the calls to the EstimatedPartitionCount metric (eg.
> org.apache.cassandra.metrics:type=Table,keyspace=testks,scope=testcf,name=EstimatedPartitionCount)
> Using the following patched function:
> {code:java}
> public RefViewFragment selectAndReference(Function<View,
> Iterable<SSTableReader>> filter)
> {
> long failingSince = -1L;
> boolean first = true;
> while (true)
> {
> ViewFragment view = select(filter);
> Refs<SSTableReader> refs = Refs.tryRef(view.sstables);
> if (refs != null)
> return new RefViewFragment(view.sstables, view.memtables,
> refs);
> if (failingSince <= 0)
> {
> failingSince = System.nanoTime();
> }
> else if (System.nanoTime() - failingSince >
> TimeUnit.MILLISECONDS.toNanos(100))
> {
> List<SSTableReader> released = new ArrayList<>();
> for (SSTableReader reader : view.sstables)
> if (reader.selfRef().globalCount() == 0)
> released.add(reader);
> NoSpamLogger.log(logger, NoSpamLogger.Level.WARN, 1,
> TimeUnit.SECONDS,
> "Spinning trying to capture readers {},
> released: {}, ", view.sstables, released);
> if (first)
> {
> first = false;
> try {
> throw new RuntimeException("Spinning trying to
> capture readers");
> } catch (Exception e) {
> logger.warn("Spin lock stacktrace", e);
> }
> }
> failingSince = System.nanoTime();
> }
> }
> }
> {code}
> Digging into this code I found it will fail if any of the sstables are in
> released state (ie. reader.selfRef().globalCount() == 0).
> See the extract.log for an example of one of these spin lock occurrences.
> Sometimes these spin locks last over 5 minutes. Across the worst cluster with
> this issue, I ran a log processing script that everytime the 'Spinning trying
> to capture readers' was different to previous one it would output if the
> released tables were in Compacting state. Every single occurrence has it spin
> locking with released listing a sstable that is compacting.
> In the extract.log example its spin locking saying that nb-320533-big-Data.db
> has been released. But you can see prior to it spinning that sstable is
> involved in a compaction. The compaction completes at 01:03:36 and the
> spinning stops. nb-320533-big-Data.db is deleted at 01:03:49 along with the
> other 9 sstables involved in the compaction.
>
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscr...@cassandra.apache.org
For additional commands, e-mail: commits-h...@cassandra.apache.org
