Temporary Resync Tuning

There may be a situation where you need to temporarily speed up the resync of a single resources in LINSTOR. The example below is from a kubernetes environment, but the same steps can be applied to any LINSTOR cluster

Below is an example of me changing the setting for the pvc-78df9b41-3064-4666-b828-679779397c4b resource. You'll likely want to do the below on all LINSTOR pods where a replica is present, but I usually start with the node that is the SyncTarget.

  • Connect to the LINSTOR Satellite pod (ns-node pod) of the SyncTarget
# kubectl exec -it linstor-op-ns-node-44p64 -- /bin/sh
  • Check the current setting with 'drbdsetup show'. You can use this afterwards to also confirm the changes.
# drbdsetup show pvc-78df9b41-3064-4666-b828-679779397c4b --show-default
  • Change the max-buffers value. 20k is a good starting value, but we've seen good results with anything between 20k-80k.
# drbdadm net-options pvc-78df9b41-3064-4666-b828-679779397c4b --max-buffers=20k
  • Change the c-fill-target to 1M.
drbdadm peer-device-options pvc-78df9b41-3064-4666-b828-679779397c4b --c-fill-target=1M
  • Connect to the satellite pod of the other peers and repeat the above as needed.
These setting are temporary, and will be cleared if the resource is ever downed/upped, or the node is restarted. This should help to increase the resync speeds if there is available disk and network throughput. Once the sync is complete you can revert to default settings with the below command. Again, you'll want to run this on all nodes where the resource is present:
# drbdadm adjust pvc-78df9b41-3064-4666-b828-679779397c4b