[ https://issues.apache.org/jira/browse/OAK-6984?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Michael Dürig updated OAK-6984:
-------------------------------
Issue Type: Bug (was: Improvement)
> High read IO in compaction retry cycles
> ---------------------------------------
>
> Key: OAK-6984
> URL: https://issues.apache.org/jira/browse/OAK-6984
> Project: Jackrabbit Oak
> Issue Type: Bug
> Components: segment-tar
> Reporter: Michael Dürig
> Assignee: Michael Dürig
> Labels: performance
> Fix For: 1.8, 1.7.13
>
> Attachments: compactioncycles.m, cpu.png, cycle-times.png, cycles.png, extract.sc,
gcmetrics-15.png, gcmetrics-20.png, gcmetrics.m, readwrite.m, readwrite.png
>
>
> We have seen unusual high read IO in compaction retry cycles in our longevity tests at
Adobe.
> !cpu.png|width=1000!
> Above picture shows the CPU utilisation during an online compaction run, which starts
at 03:00. At about 03:22 CPU user time drops and CPU waiting for IO time increases. This point
in time coincides with the start of the first retry cycle of compaction.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
|