Open PrasadDesala opened 5 years ago
var-run-glusterd2-bricks-pvc-381a5faf-0e6e-11e9-af0b-525400f94cb8-subvol1-brick3-brick.log.gz var-run-glusterd2-bricks-pvc-381a5faf-0e6e-11e9-af0b-525400f94cb8-subvol1-brick2-brick.log.gz var-run-glusterd2-bricks-pvc-381a5faf-0e6e-11e9-af0b-525400f94cb8-subvol1-brick1-brick.log.gz kube3-glusterd2.log.gz kube2-glusterd2.log.gz kube1-glusterd2.log.gz
Observed behavior
Seeing stale sockets fds under brick process with brick-mux enabled. If I restart the glusterd2 service then the sockets count are getting increased with a count of number of volumes.
Right now on my system I have 100 pvc and I am seeing 419 brick sockets on a node. [root@gluster-kube1-0 11720]# ll /proc/$(pgrep glusterfsd)/fd | wc -l 419
Expected/desired behavior
Stale socket fds should be cleaned.
Details on how to reproduce (minimal and precise)
1) Create a 3 node gcs system using valgrind. 2) Create 100 pvc. 3) Enabled brick-mux. 4) Stop/start the volume. 5) Check the brick process sockets using "ll /proc/$(pgrep glusterfsd)/fd"
Information about the environment: