[LU-7735] No way to revert the "no_primnode" flag on targets Created: 03/Feb/16 Updated: 17/Dec/16 Resolved: 17/Dec/16 |
|
| Status: | Resolved |
| Project: | Lustre |
| Component/s: | None |
| Affects Version/s: | Lustre 2.6.0, Lustre 2.7.0, Lustre 2.5.3, Lustre 2.8.0 |
| Fix Version/s: | Lustre 2.10.0 |
| Type: | Bug | Priority: | Trivial |
| Reporter: | Sebastien Piechurski | Assignee: | Nathaniel Clark |
| Resolution: | Fixed | Votes: | 0 |
| Labels: | patch | ||
| Severity: | 4 |
| Rank (Obsolete): | 9223372036854775807 |
| Description |
|
When formatting a target or modifying an existing one with the parameter "--servicenode", the flag no_primnode is set on the target, modifying the way the target is then registered on the MGS to avoid adding the NID of the node actually registering and let only the nids given with --servicenode parameters. However, if someone wants to revert to the older behavior with the "--failnode" parameter, the no_primnode flag is never unset, and the primary node NID is never added to the mountdata. A trivial fix is to unset the flag when using the "--failnode" parameter. I am currently unable to sign in to gerrit, so here is the patch inline: diff --git a/lustre/utils/mkfs_lustre.c b/lustre/utils/mkfs_lustre.c
index 8bf063a..f2f6f53 100644
--- a/lustre/utils/mkfs_lustre.c
+++ b/lustre/utils/mkfs_lustre.c
@@ -394,6 +394,7 @@ int parse_opts(int argc, char *const argv[], struct mkfs_opts *mop,
/* Must update the mgs logs */
mop->mo_ldd.ldd_flags |= LDD_F_UPDATE;
if (opt == 'f') {
+ mop->mo_ldd.ldd_flags &= ~LDD_F_NO_PRIMNODE;
failnode_set = 1;
} else {
mop->mo_ldd.ldd_flags |= LDD_F_NO_PRIMNODE;
|
| Comments |
| Comment by Peter Jones [ 03/Feb/16 ] |
|
Sebastien Is your intention to make this a Sev 1 - which indicates a production site out of service? Reading the description it does not sound that way, but, if so, which site is affected? Thanks Peter |
| Comment by Sebastien Piechurski [ 03/Feb/16 ] |
|
Oops, sorry Peter, I inverted the meaning of severity. |
| Comment by Peter Jones [ 03/Feb/16 ] |
|
|
| Comment by Peter Jones [ 03/Feb/16 ] |
|
Nathaniel Could you please look into this proposed change and see why this was changed in the first place? THanks Peter |
| Comment by Gerrit Updater [ 04/Feb/16 ] |
|
Nathaniel Clark (nathaniel.l.clark@intel.com) uploaded a new patch: http://review.whamcloud.com/18304 |
| Comment by Nathaniel Clark [ 04/Feb/16 ] |
|
Going back to the old behaviour seems to have been overlooked when servicenode was added. I'll add this as a patch. |
| Comment by Gerrit Updater [ 17/Dec/16 ] |
|
Oleg Drokin (oleg.drokin@intel.com) merged in patch https://review.whamcloud.com/18304/ |
| Comment by Peter Jones [ 17/Dec/16 ] |
|
Landed for 2.10 |