[lustre-devel] Cannot find documentation or comment blocks specifically defining lnet metrics

[Andreas Dilger]

Did you try wiki.whamcloud.com<http://wiki.whamcloud.com>?  Amir used to like creating pages for LNet there, so there might be something.

Otherwise, I can't really help with the LNet questions.

On Jan 22, 2024, at 12:35, Jeff Johnson <jeff.johnson at aeoncomputing.com<mailto:jeff.johnson at aeoncomputing.com>> wrote:

Greetings all...

I'm modifying a Lustre metrics reporting tool and incorporating lnet metrics into the data gathered. I dug through all the documentation I could find and looked through source for comment blocks and didn't find anything definitive. I'm hoping someone here could clarify. While some metrics are obvious (*_drop) others I would rather not develop to based on assumptions.

msgs_alloc
msgs_max
rst_alloc
send_count
recv_count
route_count
drop_count
send_length
recv_length
route_length
drop_length

If someone can point me to the document that clearly defines the above I'd appreciate it, so far I can't find any. As I said I don't want to develop based on assumptions.

For example, when sampled over a short period of time does...

((X_length / X_count) / # samples) = avg rpc size of X type?

Send, Recv, Route and Drop are fairly clear. Length and Count I am assuming are just that, n bytes and IOs. msgs_alloc, msgs_max, rst_alloc are less obvious.

Thanks!

--Jeff

_______________________________________________
lustre-devel mailing list
lustre-devel at lists.lustre.org<mailto:lustre-devel at lists.lustre.org>
http://lists.lustre.org/listinfo.cgi/lustre-devel-lustre.org

Cheers, Andreas
--
Andreas Dilger
Lustre Principal Architect
Whamcloud







-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.lustre.org/pipermail/lustre-devel-lustre.org/attachments/20240122/5e49ea0e/attachment.htm>