Please don't reply to lustre-devel. Instead, comment in Bugzilla by using the
following link:
https://bugzilla.lustre.org/show_bug.cgi?id=11330
What |Removed |Added
----------------------------------------------------------------------------
Attachment #9251|review?([EMAIL PROTECTED])|review+
Flag| |
(From update of attachment 9251)
>@@ -698,6 +699,21 @@ static int ost_brw_read(struct ptlrpc_re
> if (rc != 0)
> GOTO(out_bulk, rc);
>
>+ /*
>+ * If getting the lock took more time than
>+ * client was willing to wait, drop it. b=11330
>+ */
>+ if (cfs_time_current_sec() > req->rq_arrival_time.tv_sec +
>obd_timeout ||
>+ OBD_FAIL_CHECK(OBD_FAIL_OST_DROP_REQ)) {
>+ no_reply = 1;
>+ CERROR("Dropping timed-out read request from %s "
>+ "because taking the lock: "LPX64" took %ld seconds.\n",
>+ libcfs_id2str(req->rq_peer),
>+ lockh.cookie,
>+ cfs_time_current_sec() - req->rq_arrival_time.tv_sec);
minor, but still pretty desirable: no point in printing lock cookie, it does
not tell us anything useful (without having debug lock.
Better to print object id ioo->ioo_id (so it will read like "... taking the
lock on object NUMBER")
>@@ -925,6 +944,21 @@ static int ost_brw_write(struct ptlrpc_r
> if (rc != 0)
> GOTO(out_bulk, rc);
>
>+ /*
>+ * If getting the lock took more time than
>+ * client was willing to wait, drop it. b=11330
>+ */
>+ if (cfs_time_current_sec() > req->rq_arrival_time.tv_sec +
>obd_timeout ||
>+ OBD_FAIL_CHECK(OBD_FAIL_OST_DROP_REQ)) {
>+ no_reply = 1;
>+ CERROR("Dropping timed-out write request from %s "
>+ "because taking the lock: "LPX64" took %ld seconds.\n",
>+ libcfs_id2str(req->rq_peer),
>+ lockh.cookie,
>+ cfs_time_current_sec() - req->rq_arrival_time.tv_sec);
Same as above.
Otherwise is good.
Needs to be landed on b1_4 and b1_5
_______________________________________________
Lustre-devel mailing list
[email protected]
https://mail.clusterfs.com/mailman/listinfo/lustre-devel
