On 12/8/2016 1:06 AM, forest_soup wrote: > As you can see in the gc log, the long GC pause is not a full GC. It's a > young generation GC instead. > In our case, full gc is fast and young gc got some long stw pause. > Do you have any comments on that, as we usually believe full gc may cause > longer pause, but young generation should be ok?
While full GC is *typically* where long pauses happen, it can happen with *any* collection. The startup script in Solr 5.0 and later comes with GC tuning, so you probably should not be messing with it at all. If you feel that you must change the GC tuning to G1GC, perhaps you should try these settings from my personal wiki page: https://wiki.apache.org/solr/ShawnHeisey#Current_experiments Also, I would strongly recommend that you drop your max heap to 31GB instead of 32GB to change the pointer size, and that you investigate whether you need a heap that large *at all*. The numbers I saw in your log on the Jira issue did not indicate the need for that much heap memory. After the long collection, your heap usage was only 3450MB. In addition to reducing your heap size to a level that's more appropriate to your Solr's needs, I would suggest that you try the GC tuning that Solr has out of the box, which uses CMS. Thanks, Shawn
