Re: Random IO errors on nfs clients running linux > 4.20

[Alberto Gonzalez Iniesta <alberto.gonzalez@udima.es>]

On Mon, Oct 26, 2020 at 10:14:52AM -0400, J. Bruce Fields wrote:
> Sorry, I just don't know what this is off the top of my head.  If I had
> the time, stuff I might try:
> 
> 	- look at the wire traffic with wireshark: try to figure out
> 	  which operation this is happening on, and if there's anything
> 	  unusual about the reply.  May be difficult if a lot of traffic
> 	  is required to reproduce.
> 	- if you're using krb5, try without just to see if that makes a
> 	  difference.
	
Using "sec=krb5p" here... And with lots of active clients. :-/

Oh! Could I add another entry in /etc/exports with krb5 without afecting
current krb5p clients?

> 	- look at history: gitk v4.19..v4.20.17 fs/nfs net/sunrpc, see
> 	  if anything looks relevant; or even just try a git bisect
> 	  (difficult again given the intermittent failure).
> 	- trace through the code to work out where call_decode might be
> 	  returning -EIO, try to narrow it down with tracepoints or even
> 	  just debugging printk's.

I'm afraid I don't have the skills to carry out those tasks, haven't
coded in C for decades, not to mention kernel/nfs hacking.
I guess we'll have to wait until someone else hits this too, or we
update the server and (hopefully/miraculously) fixes it self :-)

Thanks a lot,

Alberto

-- 
Alberto González Iniesta             | Universidad a Distancia
alberto.gonzalez@udima.es            | de Madrid