AW: Wheezy "NFSv4 callback reply buffer overflowed"
Hello Javier >>> Did you try disabling nfs.v4.read_delegation? >>> >> No, since we only recently changed from NFSv3 to NFSv4 and with v3 were >> using mount option nocto on 99% of the mounts for performance. We have >> currently no idea what implications we would have if we turn off read >> delegation. >> Is there any way to estimate/measure the performance improvements achieved >> by using read delegation as in our current setup? >I have no idea, I'm suggesting such solution because that is the >response on Redhat customer portal (I'm not sure what could I post >from that response, I have a account there)) >(https://access.redhat.com/solutions/728943) Disabled read delegation on Sunday - had to reenable it now since the NetApp NFS server was doing 150% more NFS operations than before and the CPU was going over our limits. So we might just live with that warning then. Strange though: NFSv4 is now around 12 years old and still not properly implemented on the clients as it seems... Best regards Rainer
Re: Wheezy NFSv4 callback reply buffer overflowed
Hi, same problem here: Wheezy Linux 3.2.0-4-amd64 #1 SMP Debian 3.2.68-1+deb7u3 x86_64 GNU/Linux running NFSv4 against a NetApp FAS2240 (ONTAP 8.2.3P2 7-Mode) Multiple diskless NFS rooted systems reporting that warning: infra03-a/syslog:Aug 27 14:22:16 infra03-a kernel: [164600.492851] NFSv4 callback reply buffer overflowed! jira01/syslog:Aug 26 09:54:14 jira01 kernel: [64595.755284] NFSv4 callback reply buffer overflowed! nodejs01/syslog:Aug 26 21:11:32 nodejs01 kernel: [104524.818241] NFSv4 callback reply buffer overflowed! NFS config of NetApp system: nfs.v4.acl.enableoff(value might be overwritten in takeover) nfs.v4.enableon (value might be overwritten in takeover) nfs.v4.id.allow_numerics on nfs.v4.id.domain nfs.v4.read_delegation on (value might be overwritten in takeover) nfs.v4.write_delegation off(value might be overwritten in takeover) Mount option NFSv4 of client: rw,noatime,nodiratime,vers=4,rsize=65536,wsize=65536,namlen=255,hard,nocto,proto=tcp,port=0,timeo=100,retrans=360,sec=sys ,minorversion=0,local_lock=none Best regards Rainer
Re: Wheezy NFSv4 callback reply buffer overflowed
Hi Javier, nfs.v4.read_delegation on (value might be overwritten in takeover) nfs.v4.write_delegation off(value might be overwritten in takeover) Did you try disabling nfs.v4.read_delegation? No, since we only recently changed from NFSv3 to NFSv4 and with v3 were using mount option nocto on 99% of the mounts for performance. We have currently no idea what implications we would have if we turn off read delegation. Is there any way to estimate/measure the performance improvements achieved by using read delegation as in our current setup? From looking through the source the message seems to be just a warning, right? Thanks Rainer