isi_sysctl_cluster

I forgot to mention that these sysctls don't show up when using sysctl.... so....

AFAIK isi_sysctl_cluster can be used to modify any settable sysctl value, so it's as "dangerous" as sysctl itself (you certainly have seen the warnings in the related KB articles).

The complete list: sysctl -a -d   (never run sysctl -a without -d on a production cluster)

The list of sysctl params and values that so far have been modified
by isi_sysctl_cluster on a cluster is recorded in

/ifs/.ifsvar/etc/mcp/override/sysctl.conf

hth

-- Peter

Amazing   

Sometimes what I recall to be a sysctl turns out to appear as a isi_gconfig param...

Seems like a glitch in the Rel Notes to me...  these sysctls don't exist in 8.0.0.0 and I think you already have looked for possible (renamed) equivalents.

The 8.0.0.0  Tech Specs say:

So I'd say the relevant params are indeed via isi_gconfig (from my 8.0.0.0 virtual node):

registry.Services.lwio.Parameters.Drivers.nfs.WorkersToCoresRatio (uint32) = 4

registry.Services.lwio.Parameters.Drivers.onefs_nfs.OnefsMaxIoWorkers (uint32) = 32

registry.Services.lwio.Parameters.Drivers.onefs_nfs.OnefsMaxWorkers (uint32) = 128

That might help when talking with support; let us know when you got the real stuff.

-- Peter

But would isi_sysctl_cluster list isi_gconfig params?

I'm getting these straight from the OneFS 8.0.0.0 release notes (and they do produce valid output on my 7.2.1 cluster):

The default number of NFS server threads was changed to address a potential issue in which the NFS server monopolizes node resources. NFS performance might be lower than expected. Workaround: Adjust the number of nfsd threads by running the following commands. Modify the minimum number of threads by running the following command, where is an integer: isi_sysctl_cluster vfs.nfsrv.rpc.threads_min= Modify the maximum number of threads by running the following command, where is an integer: isi_sysctl_cluster vfs.nfsrv.rpc.threads_max= We recommend that you set threads_min and threads_max to the same value. Increasing the number of threads can improve performance, but can also cause node stability issues

Hello Carlilek,

I just spoke with one of our NFS coaches on the break fix side of the house and he confirmed that since NFS was moved to userspace (7.2.x.x) those values are now dynamic. They used to be configurable pre 7.2 but it is now in the gconfig settings as a result. again these should be dynamic at this point in OneFS and I am not aware of these being changed after 7.2

Yes, I know that, but previous to this, no one was ever able to even tell me what the range was, much less how to check it on my cluster. What I am ultimately after is how to find what the current threadcount is.

>  find what the current threadcount is.

# isi_for_array -s 'ps o nlwp,pid,command  $(pgrep -f lw-container\ nfs)'

isilon-1:  NLWP   PID COMMAND

isilon-1:    56 18955 lw-container nfs (nfs)

isilon-2:  NLWP   PID COMMAND

isilon-2:    72 62722 lw-container nfs (nfs)

isilon-3:  NLWP   PID COMMAND

isilon-3:    58 54649 lw-container nfs (nfs)

isilon-4:  NLWP   PID COMMAND

isilon-4:    52 41660 lw-container nfs (nfs)

isilon-5:  NLWP   PID COMMAND

isilon-5:    62 10994 lw-container nfs (nfs)

isilon-6:  NLWP   PID COMMAND

isilon-6:   114 20501 lw-container nfs (nfs)

isilon-7:  NLWP   PID COMMAND

isilon-7:   110 28173 lw-container nfs (nfs)

isilon-8:  NLWP   PID COMMAND

isilon-8:   107  2625 lw-container nfs (nfs)

isilon-9:  NLWP   PID COMMAND

isilon-9:    53 80716 lw-container nfs (nfs)

isilon-10: NLWP   PID COMMAND

isilon-10:  140 34265 lw-container nfs (nfs)

isilon-11: NLWP   PID COMMAND

isilon-11:   11 54905 lw-container nfs (nfs)

NLWP = num of lightweight processes aka threads

Thank you!

Will  the API Reference be updated for 8.0?

Cheers

-- Peter