"i usually see lots of compation work after I restart hbase cluster after running for some time."

[Michael Stack <st...@duboce.net>]

(Answering an IRC question up here so Rong-en, the asker, sees response)

In the log it will say what kind of compaction it is, or explicitly it 
logs when its a 'major' compaction.   Are the compactions you are seeing 
'major' compactions?  My guess is they are.  Major compactions happen on 
a period (Default every 24 hours).  But we don't persist the time of the 
last major compaction and I'm guessing that because we don't, every 
restart runs one.  Can you confirm Rong-en?  If this is the case, its a 
bug.  Please file a JIRA.  We need to fix it.

St.Ack

Re: "i usually see lots of compation work after I restart hbase cluster after running for some time."

[Rong-en Fan <gr...@gmail.com>]

The log does not indicate if this is a major one, one has to enable
debug log to see that. I will set up another cluster for test this. I
also check HRS code in branch-0.18 and agree that we didn't persist on
the major compaction period (default 1 day). We only request the
compact at 1) region open and 2) region split/merge, at which we check
the last major compaction time.

After I confirm the compaction I see at start up are all majors, I
will file an issue.

Thanks, stack!

Regards,
Rong-En Fan

On Mon, Oct 20, 2008 at 2:01 AM, Michael Stack <st...@duboce.net> wrote:
> (Answering an IRC question up here so Rong-en, the asker, sees response)
>
> In the log it will say what kind of compaction it is, or explicitly it logs
> when its a 'major' compaction.   Are the compactions you are seeing 'major'
> compactions?  My guess is they are.  Major compactions happen on a period
> (Default every 24 hours).  But we don't persist the time of the last major
> compaction and I'm guessing that because we don't, every restart runs one.
>  Can you confirm Rong-en?  If this is the case, its a bug.  Please file a
> JIRA.  We need to fix it.
>
> St.Ack
>