Description of problem: Registered close to 2000 content hosts to satelitte server. tried multiple times concurrent connections as well ( 50, 100, 150, 200) Noticed max number of java connections opened. lsof | cut -d ' ' -f 1 | sort | uniq -c | sort -n | tail 364 sshd 944 qdrouterd 1052 Passenger 2305 postgres 2523 mongod 4340 httpd 7898 ruby 8767 celery 11574 qpidd 167041 java [root@satserver elasticsearch]# ulimit -a | grep open open files (-n) 1024 [root@satserver elasticsearch]# Version-Release number of selected component (if applicable): How reproducible: Steps to Reproduce: 1. Register content hosts in scale 2. 3. Actual results: Too many open connections Expected results: Expected to be lesser open connections Additional info:
There is a lots of connections even when you are not at scale. I have Sat6 running for few days and have synced RHEL6 + RH Common repos and registered ~10 systems and I see this: # lsof | cut -d ' ' -f 1 | sort | uniq -c | sort -n | tail 344 tuned 420 named 636 Passenger 720 qpidd 1950 postgres 1975 mongod 3521 ruby 4070 httpd 6489 celery 56592 java # lsof | grep elastic | wc -l 40320
Moving 6.2 bugs out to sat-backlog.
This bug appears to be related to elasticsearch, which has been removed in 6.2. Marking as CLOSED/CURRENTRELEASE.