[LU-14855] nodemap_add_range failed: MGS: OBD_IOC_NODEMAP command CE042 for ... : rc = -12 Created: 16/Jul/21  Updated: 19/Jul/21  Resolved: 19/Jul/21

Status: Resolved
Project: Lustre
Component/s: None
Affects Version/s: None
Fix Version/s: None

Type: Bug Priority: Major
Reporter: Matt Rásó-Barnett (Inactive) Assignee: WC Triage
Resolution: Duplicate Votes: 0
Labels: None

Issue Links:
Duplicate
duplicates LU-14802 MGS configuration problems - cannot a... Resolved
Severity: 2
Rank (Obsolete): 9223372036854775807

 Description   

Hello,

I attempted to add a new range to a nodemap that we use in production here, and got the following error:

[root@rds-mds10 ~]# lctl nodemap_add_range --name csd3 --range 10.43.240.[198-199]@tcp2                                                                                                                                                                        
error: invalid ioctl: 000ce042 errno: 12 with rc=-1
error: nodemap_add_range: cannot add range '10.43.240.[198-199]@tcp2' to nodemap 'csd3': rc = -1

With the following in syslog:

Jul 16 15:27:08 rds-mds10 kernel: LustreError: 111140:0:(nodemap_handler.c:812:nodemap_add_range_helper()) cannot insert nodemap range into 'csd3': rc = -17
Jul 16 15:27:08 rds-mds10 kernel: LustreError: 111140:0:(mgs_handler.c:892:mgs_iocontrol_nodemap()) MGS: OBD_IOC_NODEMAP command CE042 for csd3: rc = -12 

The environment of this server is:

[root@rds-mds10 ~]# lfs --version
lfs 2.12.5
[root@rds-mds10 ~]# uname -r
3.10.0-1127.8.2.el7_lustre.x86_64

Do those errors give any clue as to what the cause of the problem is here?

Thanks,
Matt



 Comments   
Comment by John Hammond [ 16/Jul/21 ]

Does this range already exist in the nodemap?

Comment by Andreas Dilger [ 16/Jul/21 ]

Just looking at nodemap_add_range_helper() it appears this error is generated from range_insert() returning -EEXIST = -17. Are you sure this range is not already included in the csd3 nodemap?

Comment by Matt Rásó-Barnett (Inactive) [ 16/Jul/21 ]

Weirdly, yes, it is showing up when viewing the ranges from this server, which is in a HA pair and is mounting only the MGS currently:

[root@rds-mds10 ~]# lctl get_param nodemap.csd3.ranges | grep 10.43.240
 { id: 88, start_nid: 10.43.240.198@tcp2, end_nid: 10.43.240.199@tcp2 },

but is not visible from the partner server, currently mounting the MDTs for this filesystem (and was where I was testing this from initially).

[root@rds-mds9 ~]# lctl get_param nodemap.csd3.ranges | grep 10.43.240.198

and this nodemap has a fileset configured, and the clients whose IPs I'm trying to add here, cannot mount the filesystem.

So I guess this is some kind of communication problem with the MGS, possibly related to my other issue mentioned in LU-14802

For what it's worth I've tried simply restarting the MGS itself, but that didn't help things. I could try fail it over to the other server, but due to the issues I was having in LU-14802, I've kept the MGT device on a separate server so it can be restarted without impacting the MDT devices.

Thanks for the quick replies,
Matt

Comment by Peter Jones [ 19/Jul/21 ]

Let's focus on LU-14802

Generated at Sat Feb 10 03:13:22 UTC 2024 using Jira 9.4.14#940014-sha1:734e6822bbf0d45eff9af51f82432957f73aa32c.