[
https://issues.apache.org/jira/browse/HBASE-2536?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12868021#action_12868021
]
Jonathan Gray commented on HBASE-2536:
--------------------------------------
Looks like the blockStoreFiles threshold is looked at on a per-store basis.
>From MemStoreFlusher:
{code}
for (Store hstore: region.stores.values()) {
if (hstore.getStorefilesCount() > this.blockingStoreFilesNumber) {
{code}
where
{code}
this.blockingStoreFilesNumber =
conf.getInt("hbase.hstore.blockingStoreFiles", -1);
if (this.blockingStoreFilesNumber == -1) {
this.blockingStoreFilesNumber = 1 +
conf.getInt("hbase.hstore.compactionThreshold", 3);
}
{code}
So this shouldn't be an issue.
> blocking store files count should be per-store, not per-region
> --------------------------------------------------------------
>
> Key: HBASE-2536
> URL: https://issues.apache.org/jira/browse/HBASE-2536
> Project: Hadoop HBase
> Issue Type: Bug
> Components: regionserver
> Affects Versions: 0.20.4
> Reporter: Todd Lipcon
> Priority: Critical
>
> The hbase.hstore.blockingStoreFiles threshold is currently across the entire
> region, which means that if the number of column families multiplied by
> hbase.hstore.compactionThreshold is greater than the threshold, writes will
> freeze up, since we neither flush nor compact.
> We should set this threshold per store, not per region, probably.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
