cassandra-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Kai Wang (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (CASSANDRA-6149) OOM in Cassandra 2.0.1
Date Sun, 06 Oct 2013 12:48:42 GMT

    [ https://issues.apache.org/jira/browse/CASSANDRA-6149?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13787587#comment-13787587
] 

Kai Wang commented on CASSANDRA-6149:
-------------------------------------

MAT shows "One instance of "com.google.common.cache.LocalCache" loaded by "sun.misc.Launcher$AppClassLoader
@ 0xd03530b0" occupies 942,959,256 (91.01%) bytes."

Another thing I notice is this OOM accompanoes with a LOT of "Unable to delete ... (it will
be removed on server restart; we'll also retry after GC)" errors. I understand it might be
OK on Windows. But the amount of such errors I see with 1e0d9513b748fae4ec0737283da71c65e9272102
seem unusual. In fact I don't see a single such error   prior to 1e0d9513b748fae4ec0737283da71c65e9272102.

> OOM in Cassandra 2.0.1
> ----------------------
>
>                 Key: CASSANDRA-6149
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-6149
>             Project: Cassandra
>          Issue Type: Bug
>          Components: Core
>         Environment: Windows 7 64 bit. Java 64-bit 1.7.0_25. Cassandra 2.0.1
>            Reporter: Kai Wang
>              Labels: newbie
>
> I have a program to stress test Cassandra. What it does is remove/insert rows with a
small set of row keys as fast as possible. Two CFs are involved. When I test against C* 1.2.3
with default configurations, it ran for 24 hours and C* doesn't having any issue. However
after I upgraded to C* 2.0.1, C* crashes on OOM within 1-2 minutes. I can consistently reproduce
this.
> I built C* from the source and found out the last good changeset is cfa097cdd5e28d7fe8204248e246a1fae226d2c0.
As soon as I include the next changeset 1e0d9513b748fae4ec0737283da71c65e9272102, C* starts
to crash. What's interesting is although it seems the change was reverted by fc1a7206fe15882fd64e7ba8eb68ba9dc320275f.
C* built from fc1a7206fe15882fd64e7ba8eb68ba9dc320275f has the same problem - OOM within minutes.
> I didn't test against the official 2.0.0. But the C* built from 03045ca22b11b0e5fc85c4fabd83ce6121b5709b
seems OK. I assume that's what 2.0.0 is.
> I use default configurations in all cases. I didn't tune anything.



--
This message was sent by Atlassian JIRA
(v6.1#6144)

Mime
View raw message