On Fri, Jun 19, 2015 at 4:17 AM, Jeff Layton <jeff.layton@xxxxxxxxxxxxxxx> wrote: > On Tue, 16 Jun 2015 22:47:21 +0800 > Peng Tao <tao.peng@xxxxxxxxxxxxxxx> wrote: > >> Hi all, >> >> The patchsets add LAYOUTSTATS support to flexfiles. LAYOUTSTATS are sent >> every minute if IO is still happening upon a file. >> >> One limitation is that at most 4 LAYOUTSTATS calls are permitted in a compound. >> Had to send multiple LAYOUTSTATS operations per compound because OP_LAYOUTSTATS >> requires stateid and deviceid as its arguments, which makes it a per-file per-deviceid >> call. >> > > So what happens if there are more than 4 mirrors? Do you send more > LAYOUTSTATS rpcs or do the other ones get left out? I couldn't quite > tell that from looking over the code... > Only the first four mirrors' stats are sent out. The rest are left out. It is in ff_layout_prepare_layoutstats(). >> Cheers, >> Tao >> >> Peng Tao (8): >> pNFS: fill in nfs42_layoutstat_ops >> pnfs: add pnfs_report_layoutstat helper function >> pNFS/flexfiles: track when layout is first used >> pnfs/flexfiles: add ff_layout_prepare_layoutstats >> pnfs/flexfiles: encode LAYOUTSTATS flexfiles specific data >> pnfs/flexfiles: reset IO statistics upon LAYOUTSTATS success >> nfs42: serialize LAYOUTSTATS calls of the same file >> pnfs/flexfiles: report layoutstat regularly >> >> Trond Myklebust (3): >> NFSv.2/pnfs Add a LAYOUTSTATS rpc function >> pNFS/flexfiles: Remove unused struct members user_name, group_name >> pNFS/flexfiles: add layoutstats tracking >> >> fs/nfs/flexfilelayout/flexfilelayout.c | 448 ++++++++++++++++++++++++++++++++- >> fs/nfs/flexfilelayout/flexfilelayout.h | 30 ++- >> fs/nfs/nfs42.h | 7 +- >> fs/nfs/nfs42proc.c | 81 ++++++ >> fs/nfs/nfs42xdr.c | 122 +++++++++ >> fs/nfs/nfs4_fs.h | 1 + >> fs/nfs/nfs4proc.c | 4 +- >> fs/nfs/nfs4xdr.c | 1 + >> fs/nfs/pnfs.c | 56 +++++ >> fs/nfs/pnfs.h | 3 + >> include/linux/nfs4.h | 1 + >> include/linux/nfs_fs.h | 1 + >> include/linux/nfs_xdr.h | 43 ++++ >> 13 files changed, 782 insertions(+), 16 deletions(-) >> > > Nice work, Tao. This looks pretty good to me. I had a couple of > nitpicky comments, but that's stuff that can wait unless you need to > respin for other reasons. > > You can add my: > > Reviewed-by: Jeff Layton <jeff.layton@xxxxxxxxxxxxxxx> Thanks Jeff! Cheers, Tao -- To unsubscribe from this list: send the line "unsubscribe linux-nfs" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
