[
https://issues.apache.org/jira/browse/CASSANDRA-12730?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15640570#comment-15640570
]
Benjamin Roth commented on CASSANDRA-12730:
-------------------------------------------
After having thought a bit about that issue, I come up with the following
proposal:
Given that it is necessary to flush memtables for streaming to have immuntable
data and a consistent state to stream, we only need to flush the memtable if it
contains data for the token range(s) that are about to be streamed.
To achieve this, StreamSession.prepare should not always pass true as
"flushTables" but should check if the affected memtables contain partitions for
the requested range(s). To be able to determine that, the Memtable class
requires a method that takes a token range and then iterates over its own
partition keys and check if the passed token range contains that partition key.
To cap the resources that are required to determine that, it may make sense to
skip that check if a memtable contains very many partitions and a flush would
be appropriate anyway. This check shall not avoid a flush at all costs but
should avoid excessive flushes. So maybe there could be a limit of e.g. 1000
(TBD) partitions that are being checked and if the memtable contains more
partitions it is flushed without a range check. In other words, it is a more
sophisticated version of Memtable.isClean with the context a token range, like
"isCleanForRange(Range<Token>)"
The impact on existing code should be quite small and the risk of a performance
impact on production systems should be quite small as well. In fact the risk of
piling up thousands of SSTables is much higher than a calculable overhead of
some simple calculations.
How does that sound?
> Thousands of empty SSTables created during repair - TMOF death
> --------------------------------------------------------------
>
> Key: CASSANDRA-12730
> URL: https://issues.apache.org/jira/browse/CASSANDRA-12730
> Project: Cassandra
> Issue Type: Bug
> Components: Local Write-Read Paths
> Reporter: Benjamin Roth
> Priority: Critical
>
> Last night I ran a repair on a keyspace with 7 tables and 4 MVs each
> containing a few hundret million records. After a few hours a node died
> because of "too many open files".
> Normally one would just raise the limit, but: We already set this to 100k.
> The problem was that the repair created roughly over 100k SSTables for a
> certain MV. The strange thing is that these SSTables had almost no data (like
> 53bytes, 90bytes, ...). Some of them (<5%) had a few 100 KB, very few (<1%
> had normal sizes like >= few MB). I could understand, that SSTables queue up
> as they are flushed and not compacted in time but then they should have at
> least a few MB (depending on config and avail mem), right?
> Of course then the node runs out of FDs and I guess it is not a good idea to
> raise the limit even higher as I expect that this would just create even more
> empty SSTables before dying at last.
> Only 1 CF (MV) was affected. All other CFs (also MVs) behave sanely. Empty
> SSTables have been created equally over time. 100-150 every minute. Among the
> empty SSTables there are also Tables that look normal like having few MBs.
> I didn't see any errors or exceptions in the logs until TMOF occured. Just
> tons of streams due to the repair (which I actually run over cs-reaper as
> subrange, full repairs).
> After having restarted that node (and no more repair running), the number of
> SSTables went down again as they are compacted away slowly.
> According to [~zznate] this issue may relate to CASSANDRA-10342 +
> CASSANDRA-8641
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
