You can get a thread dump by calling http://localhost:8983/solr/admin/threads or by using the Admin UI.
On Wed, Apr 18, 2018 at 9:11 PM, Felix XY <felix.xy.xy...@gmail.com> wrote: > Thank you Emir, but I'm not able to make a thread dump while doing updates > because the updates are very fast again: > > > While I wrote this email my colleague was googling around. > > He found this > http://lucene.472066.n3.nabble.com/HttpSolrServer- > commit-is-taking-more-time-td4330954.html > > and my colleague changed some values > > from: > > <!-- <filterCache class="solr.FastLRUCache" > size="4096" > initialSize="1024" > autowarmCount="512"/> > --> > > to: > <filterCache class="solr.FastLRUCache" > size="512" > initialSize="512" > autowarmCount="0"/> > from: > > <!-- <queryResultCache class="solr.LRUCache" > size="16384" > initialSize="4096" > autowarmCount="1024"/> > --> > > to: > <queryResultCache class="solr.LRUCache" > size="512" > initialSize="512" > autowarmCount="0"/> > > > and it seems, that our problems are gone completely. Updates fast. > Search seems not to be much slower. > > > But I'm still curious why our problems started so suddenly and what > negative side effects these changes could have. > > Cheers > > Felix > > > > 2018-04-18 17:11 GMT+02:00 Emir Arnautović <emir.arnauto...@sematext.com>: > > > Hi Felix, > > Did you try to do thread dump while doing update. Did it show anything? > > > > Emir > > -- > > Monitoring - Log Management - Alerting - Anomaly Detection > > Solr & Elasticsearch Consulting Support Training - http://sematext.com/ > > > > > > > > > On 18 Apr 2018, at 17:06, Felix XY <felix.xy.xy...@gmail.com> wrote: > > > > > > Hello group, > > > > > > since two days we have huge problems with our solr 5.4.1 installation. > > > > > > ( yes, we have to update it. But this will not be a solution right now > ) > > > > > > All path=/select requests are still very fast. But all /update Requests > > > take >30sec up to 3 minutes. > > > > > > The index is not very big (1.000.000 documents) and its size on disk is > > > about 1GB > > > > > > The virtual server (ESX) has 8GB RAM and 8 cores. IO is good. > > > > > > solr was started with -Xms4096M -Xmx4096M > > > ( but we changed it to higher and lower values during our tests ) > > > > > > We have a lot of /select requests at the moment (10.000/Minute) but > this > > is > > > not unusual for this installation and we didn't have this update > problems > > > before. > > > > > > On another identical sleeping core on the same server, we are able to > > make > > > fast updates. We experience slow updates only on the core with high > > > selecting traffic. So it seems not to be a general problem with java, > GC, > > > .... > > > > > > We disabled all other insert/updates and we are able to reproduce this > > slow > > > update behaviour in the Solr Admin console with a single update of one > > > document. > > > > > > We are lost. > > > > > > We didn't change the Solr configuration. > > > The load seems to be not higher then during previous peaks > > > The developers didn't change anything (so they say) > > > Search is still fast. > > > > > > But single simple updates takes >30sec.... > > > > > > Any ideas about this? We tried quite a lot the last two days.... > > > > > > Cheers > > > Felix > > > > > -- Regards, Shalin Shekhar Mangar.
