Thanks a lot for the detailed explanation,
> Hi. I have enabled the mounstats, nfs and nfsd collectors for the NFS > > side metrics but there is a plethora of metrics without any proper > > documentation. Can somebody help with which metrics would give me > > the reads and writes completed on a particular mount? Like we have > > node_disk_reads_completed_total for SSD and HDDs, is there any exact > > alternative for NFS to get the IO stats? > > If you care about NFS client statistics, there is no exact analog > because NFS mount metrics are fundamentally at a different level than > disk IO statistics. Disk IO statistics are obviously at the level of > data (blocks) transfered, regardless of why they are transfered and > how much of the data will be used. NFS client mount metrics are at the > level of filesystem operations, and filesystem operations can have > unpredictable disk IO impacts. All NFS operations send data to the NFS > server and get data back from it, but some only (potentially) create > read disk IO on the server while others cause write IO. > > To put together NFS client IO statistics that are useful to you, you > will have to figure out what you care about in all of this. Some of this > will be workload dependent; for example, if your workload mostly reads > and writes a small number of big files, the filesystem level read and > write IO (which you can get numbers on) is highly predictive of disk IO > on the server and probably of performance. If your workload spends a lot > of time creating and deleting small files, looking only at the bytes > read and written from those files is probably missing a lot of server > disk IO, especially write IO. > >> Yes, I need these IOs exactly that are caused by the NFS operations. Can you give me an idea on which Prometheus metrics can give me this? > As for what the metrics mean, that is hard to summarize well here. > The 'mountstat' collector is the most detailed and more or less what > it is collecting is written up in: > > https://utcc.utoronto.ca/~cks/space/blog/linux/NFSMountstatsIndex > > >> This is quite useful. Thanks! This is from 2013 and for NFS v3, not NFS v4, but I believe that not > much has changed in this area of Linux since then and NFS v4 is pretty > similar to NFS v3. > > (Locally I have a program that significantly aggregates this information > on a per-mount basis, because we have a *lot* of NFS mounts and the raw > mountstats data for them all adds up to too many metrics for us.) > > - cks > -- You received this message because you are subscribed to the Google Groups "Prometheus Users" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To view this discussion on the web visit https://groups.google.com/d/msgid/prometheus-users/543f5ad6-5586-4e93-a0db-0ffc1f0e3108n%40googlegroups.com.
