Just to follow up, we built the patch that was made for this issue and are testing it now. The problem looks to be totally fixed and we are very happy with the improved performance and usability. Thanks everyone!
________________________________________ From: Ted Yu <[email protected]> Sent: Wednesday, October 7, 2015 3:19 PM To: [email protected] Subject: Re: Bucket cache causes region server status page to use 1GB ram bq. which seems of limited use to me Agree. Mind logging an improvement JIRA ? Cheers On Wed, Oct 7, 2015 at 12:00 PM, James Hartshorn <[email protected]> wrote: > Since we started using the HBase bucket cache I've noticed the region > server status pages not loading completely in Chrome, and causing > significant performance problems on Firefox. After some investigation I > realized the page itself is huge: > > > > wc -l HBase-Region-Server-hb22.html > 2010116 HBase-Region-Server-hb22.html > > ls -lah HBase-Region-Server-hb22.html > 32M Oct 6 19:23 HBase-Region-Server-hb22.html > > Firefox "about:memory": > ? ???1,330.18 MB (48.22%) -- top( > http://hb9.prod1.connexity.net:60030/rs-status#bc_l2, id=2010) > ? ? ???1,329.61 MB (48.20%) -- active/window( > http://hb9.prod1.connexity.net:60030/rs-status#bc_l2) > > > > It looks like it's loading a thing for each bucket, though it's not > displayed unless you click L2, which usually just hangs the browser, on > success you get a very very long list of all the buckets, their size and > occupancy, which seems of limited use to me. > > > Is there a way to stop the regionserver from sending this info for each > page load? > > > (If having a few hundred thousand buckets is completely wrong don't worry, > I'll be asking about that too soon). >
