Hi ,
I am getting out of memory issue in nfsv4 protocol in nfs ganesha verstio 2.3. IN nfsv3 protocol we did not observe such issue at that particular amount of upload. We are suspecting that packet consumption speed is slower than data reception rate. For that we analyzed nfs queue size from which worker thread picks up packets and process. We could see that there is lag in data enqueue and dequeue rate.
Also we observed that when we put delay in read from socket then this issue does not occur. We are using local and sdfs mount points.In local mount points this issue does not occur which indicates the disk write speed is primarily causing this issue.
But we could not observe any threshold value inf nfsv4 at which oom issue will occur, as compared to nfsv3.
Further oom_score does not go down even after copy is complete in both cases of local mount and sdfs mount and int both of nfsv4 and v3 protocols.
Implemeted several solution suggested like inceasing the kernel priority of reclaiming cache inode and dentries. But it could not afffect the issue.
Any suggestion would be highly appreciated.
regards
-Deepak
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi ,
I am getting out of memory issue in nfsv4 protocol in nfs ganesha verstio 2.3. IN nfsv3 protocol we did not observe such issue at that particular amount of upload. We are suspecting that packet consumption speed is slower than data reception rate. For that we analyzed nfs queue size from which worker thread picks up packets and process. We could see that there is lag in data enqueue and dequeue rate.
Also we observed that when we put delay in read from socket then this issue does not occur. We are using local and sdfs mount points.In local mount points this issue does not occur which indicates the disk write speed is primarily causing this issue.
But we could not observe any threshold value inf nfsv4 at which oom issue will occur, as compared to nfsv3.
Further oom_score does not go down even after copy is complete in both cases of local mount and sdfs mount and int both of nfsv4 and v3 protocols.
Implemeted several solution suggested like inceasing the kernel priority of reclaiming cache inode and dentries. But it could not afffect the issue.
Any suggestion would be highly appreciated.
regards
-Deepak