Why my expired translog files are not deleted?

Elastic Stack Elasticsearch
1

Hi

elasticsearch version : 6.1.3
when call IndexShard#afterWriteOperation to flush , i occured (29125)[https://github.com/elastic/elasticsearch/pull/29125]

index.translog.flush_threshold_size:5G
index.translog.retention.age:12h
index.translog.retention.size: 512M

when my .tlog's size > 5G，i ocured endless flush loop
the first .tlog file generated before 19h, total .tlog sizes have reached 6.5g

image
image.png832×314 15.8 KB

In IndexShard#afterWriteOperation shouldPeriodicallyFlush or shouldRollTranslogGeneration will call translog.trimUnreferencedReaders() , why my translog always growing has not been deleted by index.translog.retention.age or index.translog.retention.size policy ?

2

The answer to your question is the issue to which you linked, https://github.com/elastic/elasticsearch/pull/29125, which fixes a bug in the very old version you are using. You should upgrade.

3

@DavidTurner Thanks for your reply.

I see 29125 fixed the wrong logic in shouldPeriodicallyFlush() but rollTranslogGeneration() or flush() calls translog.trimUnreferencedReaders(), i want to know why
translog files has not been deleted .

    @Override
    public void rollTranslogGeneration() throws EngineException {
        try (ReleasableLock ignored = readLock.acquire()) {
            ensureOpen();
            translog.rollGeneration();
            translog.trimUnreferencedReaders();
        } catch (AlreadyClosedException e) {
            failOnTragicEvent(e);
            throw e;
        } catch (Exception e) {
            try {
                failEngine("translog trimming failed", e);
            } catch (Exception inner) {
                e.addSuppressed(inner);
            }
            throw new EngineException(shardId, "failed to roll translog", e);
        }
    }

BTW, Is there any other way to avoid this problem besides upgrading?

4

Sorry, 6.1 is well past the end of its supported life and I'm not in a position to investigate further. I don't believe there is a workaround that doesn't involve upgrading.

6

Thanks again

7

This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.