Finix1979
Finix1979
> You mention "could dramatically improve performance" - do you have benchmarks for how well it performs in some situations without and with this PR? In our production machine, we...
> If database calls fadvise for read ahead, that means application knows what they want exactly. With fadvise supported, ZFS could prefetch data accurately. The reason why we set metadata...
> Thank you. I will take a look at it.
Hi @behlendorf The reason why checksum/filetest_002_pos failed is that the cat command uses fadvise internally and the zio_checksum_verify function does not record checksum error if zio flag contains ZIO_FLAG_SPECULATIVE. if...
Thank you @nickchen-cpu If we could check if the write range is already copied or in progress(read io is in flight already) before issuing the read operation, we could avoid...
> snapshot sda1 - orig_mrf is recorded normally, tracing_mrfis swapped in snapshot sda2 - orig_mrf is set to tracing_mrf, tracing_mrf is swapped in destroy sda1 - orig_mrf (which is the...
> Indeed. Thank you @nickchen-cpu
> Thanks for catching this and posting a fix! I'd like to go with Rob's alternate change for this though which is more consistent with how we handle this kind...