Please don't reply to lustre-devel. Instead, comment in Bugzilla by using the 
following link:
https://bugzilla.lustre.org/show_bug.cgi?id=11089



cc'ing this comment to lustre-devel, referring to changing the per-export stats
on the servers to per-nid stats. Per-nid stats would be persistent across client
disconnections.

(In reply to comment #56)
> OK, new thought - forget the per-export stats, and only use the per-nid stats.
> Most of the time they will be duplicates of each other anyhow, with the main
> difference being that the per-export stats are destroyed when a client
> disconnects.
> 
> New structure:
> +- exports
>     |
>     +- nid#1
>     |   |
>     |   + stats
>     |   |
>     |   + uuids
>     |
>     +- nid#2...
>     +- clear
> 
> The "uuid"s file would list the uuids of _active_ exports, so if there was
> nothing connected from that nid at the moment, the uuids file would be empty.
> (That way people can easily see if the nid is "stale".)
> 
> The other thing we should do is have a way to clear all the stale nids, so I
> propose echoing a 0 into a new file "clear" under the exports directory.  This
> will erase any nid subdir that is not currently in use, and reset the stats of
> any nid that is in use.

_______________________________________________
Lustre-devel mailing list
[email protected]
https://mail.clusterfs.com/mailman/listinfo/lustre-devel

Reply via email to