15 second full GC pauses are far too long. I’d recommend you concentrate on sizing your hardware/heap setting to minimize the full GC pauses.
Here’s a good place to start: https://cwiki.apache.org/confluence/display/SOLR/ShawnHeisey Best, Erick > On Jan 11, 2020, at 9:14 AM, Rajdeep Sahoo <rajdeepsahoo2...@gmail.com> wrote: > > Nodes were busy with full Gc over 15 sec. > In solr console it was showing no server hosting shard,status code 500. > In solr log the same error was there i.e. solrexception: No server hosting > shard. > > Apart from this,in the zoo console it was showing recovery failing status > of the nodes. > > Thanks in advance > > On Sat, 11 Jan, 2020, 7:15 PM Erick Erickson, <erickerick...@gmail.com> > wrote: > >> You’ve provided no details, nor relayed any findings from looking >> at the Solr logs. In short, there’s not enough information here to >> provide any helpful response. >> >> Full GCs are normal, but if they’re long enough >> to exceed certain timeouts, they can trigger recoveries. Solr 4.6 had >> a number of conditions that can lead to this this, but Solr 4.6 is >> over 6 years old. There’s going to be little help available at this point. >> >> You might want to review: https://wiki.apache.org/solr/UsingMailingLists >> >> >>> On Jan 11, 2020, at 2:47 AM, Rajdeep Sahoo <rajdeepsahoo2...@gmail.com> >> wrote: >>> >>> Hi , >>> >>> Anyone Please suggest. >>> >>> On Sat, 11 Jan, 2020, 12:33 AM Rajdeep Sahoo, < >> rajdeepsahoo2...@gmail.com> >>> wrote: >>> >>>> Hi all, >>>> In production we have done the set up of solr cloud with solr version >>>> 4.6 , the set up contains 3 zookeeper and 4 shards each having three >>>> replicas, a total of 12 solr nodes. >>>> Active indexing was going on , after switching on we are experiencing a >>>> lot of issues , all the nodes stopped serving the search requests and in >>>> log it is showing recovery failing , and can see Full GC for 2 shards. >>>> After graceful restart , the recovery failing issue got resolved but >>>> currently we are using master slave with 16 slave nodes and 1 leader , >> it >>>> is working fine. >>>> Do we need to scale it up in solr cloud? Please suggest as it is a >>>> production env. , I guess all you can understand the impact of it. >>>> >>>> Thanks in advance >>>> >> >>
