[DRBD-user] DRBD 9.0.16 / PingAck did not arrive in time

[Brice CHAPPE]

Hi list,

 

I have a pb when I read from one node (secondary) to a diskless node.

I read and copy the content on this volume to the diskless node (locally).

I precise : servers have no load, network is 2x10Gb. Nothing is running in
the same time.

 

Sometimes the copy is fine. Sometimes I got on the secondary node :

Jan 15 13:54:14 os-storage-a1 kernel: drbd
CV_601b25a2-ac25-45dc-bbec-d4eee60be77d os-backup-1: sock was shut down by
peer

Jan 15 13:54:14 os-storage-a1 kernel: drbd
CV_601b25a2-ac25-45dc-bbec-d4eee60be77d os-backup-1: conn( Connected ->
BrokenPipe ) peer( Secondary -> Unknown )

Jan 15 13:54:14 os-storage-a1 kernel: drbd
CV_601b25a2-ac25-45dc-bbec-d4eee60be77d/0 drbd1088 os-backup-1: pdsk(
Diskless -> DUnknown ) repl( Established -> Off )

 

On the backup diskless node:

Jan 15 13:54:14 os-backup-1 kernel: drbd
CV_601b25a2-ac25-45dc-bbec-d4eee60be77d os-storage-a1: PingAck did not
arrive in time.

Jan 15 13:54:14 os-backup-1 kernel: drbd
CV_601b25a2-ac25-45dc-bbec-d4eee60be77d os-storage-a1: conn( Connected ->
NetworkFailure ) peer( Secondary -> Unknown )

Jan 15 13:54:14 os-backup-1 kernel: drbd
CV_601b25a2-ac25-45dc-bbec-d4eee60be77d/0 drbd1088 os-storage-a1: pdsk(
UpToDate -> DUnknown ) repl( Established -> Off )

 

I tried many times, and tried in the same time with ping (is ok), iperf
(full of 10Gb), writing with dd on volume/on local disk to put system in
load, but I can't reproduce.

 

Is there a known bug in 9.0.16 ? and 9.0.17, 18 or 19 correct it ?

 

Thanks.

 

Regards,

 

Brice

 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.linbit.com/pipermail/drbd-user/attachments/20200116/7a8951d1/attachment.htm>