ginn13

@glushakov @bykvaadm you can use `no_track_mode` feature to disable whole query text in metrics. Environment variable named PGSCV_NO_TRACK_MODE is available for this purpose.

@pawanpraka1, case is: multiple zpools per node. I would like to have ZFSVolume on any of N zpool per single node. However, StorageClass is stick to zpool name, so to...

The "slow" filer has significantly more goroutines while experiencing slow PUT requests.

SeaweedFS 3.34. The initial problem showed significant slowdown on chunkAssign Updated Seaweedfs to 3.55. The problem is gone

Well the problem returned recently Moreover, it affects only 0.1% requests. During not the top load Setup is: 3 masters, 3 filer-s3, 24 volume-servers on 6 k8s nodes. MongoDB as...

moreover, still cannot understand, why slowdown is on only 1 filer at a time, and affects only 1 bucket, though collections are well distributed across volume servers, other collection show...

@chrislusf @kmlebedev I've changed maxTimeout for Assign in master from 10 to 5 seconds https://github.com/seaweedfs/seaweedfs/blob/3.55/weed/server/master_grpc_server_volume.go#L166 And filer's chunkAssign metric went down from ~13sec to ~6 seconds I see logs in...

@kmlebedev I've already enabled log here https://github.com/seaweedfs/seaweedfs/blob/7eafa3420b8e5ae83c8873cddd03ded90a0fc921/weed/server/master_grpc_server_volume.go#L196 but not as verbose as in your PR Already catch in the master leader log Will try your PR, thanks

I've patched 3.55 with your PR to just add this Warn log. Should I update to latest version instead?

@chrislusf which limit for fd should I check? Here are open fd for master-leader ``` /data # ls /proc/1/fd/ | wc -l 102 ``` for filer-0 ``` /data # ls...