If flush interval is set to 1 hr, that wouldn't explain why you frequently encountered slow performance for deletes, right ?
Mind pastebin'ing portion of region server log ? Cheers On Jul 23, 2014, at 1:57 AM, "Li Yong" <[email protected]> wrote: > Thanks, Ted. > I see thousands upon thousands 'periodicFlusher requesting flush for > region' logs. I guess the conf > 'hbase.regionserver.optionalcacheflushinterval' effects this! We use the > default 1h. > > > Is this the reason delete delay too much? And how can I repair this > without shutdown cluster? > > > ------------------ Original ------------------ > From: "Ted Yu";<[email protected]>; > Date: Wed, Jul 23, 2014 05:41 PM > To: "[email protected]"<[email protected]>; > Cc: "user"<[email protected]>; > Subject: Re: Too many delete delay > > > > Have you checked region server logs to see if there was some clue ? > > If so, mind pastebin'ing some snippet ? > > Cheers > > On Jul 23, 2014, at 12:22 AM, "Li Yong" <[email protected]> wrote: > >> Hi all, >> I move my cluster to another cluster with CopyTable, then I found too >> many timeout, Most of them are Delete! I found the table has too many >> storefiels, so i do a major_compact, then things be better. After an hour, >> timeout again! I don't know why so many delay in delete, not put, not scan. >> Can anyone tell me, when will delete operation use too much time? Thanks! >> [ HBase version : 0.94.9 Hadoop version : 1.0.4 ]
