[LU-1168] changing allocation of ost Created: 03/Mar/12  Updated: 06/Mar/12  Resolved: 06/Mar/12

Status: Resolved
Project: Lustre
Component/s: None
Affects Version/s: Lustre 1.8.7
Fix Version/s: None

Type: Bug Priority: Blocker
Reporter: Supporto Lustre Jnet2000 (Inactive) Assignee: Cliff White (Inactive)
Resolution: Fixed Votes: 0
Labels: ldiskfs
Environment:

Red Hat Enterprise Linux 5.7 + lustre 1.8.7-wc1


Severity: 1
Epic: server
Rank (Obsolete): 6438

 Description   

We add two new OSS server and we would like to redistribute the OST on these 2 new OSS.
The old OSS are:

  • 10.121.13.31@tcp
  • 10.121.13.62@tcp
    The new OSS are:
  • 10.121.13.59@tcp
  • 10.121.13.28@tcp

I have un-mounted all the clients, and the mgt, mdt, osts. I want to change the allocation of ost06 from 10.121.13.62@tcp to 10.121.13.28@tcp. The original configuration of ost06 was:
Parameters: mgsnode=10.121.13.31@tcp failover.node=10.121.13.31@tcp ost.quota_type=ug
and the ost06 was hosted on 10.121.13.62@tcp

I have use tunefs.lustre to modify the ost configuration with this command line:
tunefs.lustre --erase-param --mgsnode=10.121.13.31@tcp --mgsnode=10.121.13.62@tcp --param="failover.node=10.121.13.59@tcp ost_quota=ug" --writeconf /dev/mapper/ost06p1
I want host the ost06 on 10.121.13.28@tcp with a failover on 10.121.13.59@tcp

When I starting up, I see in the logs that the mds try to connect to OST0006 on 10.121.13.62@tcp and give some ost_connection error.

Could you provide me the correct procedure to relocate the ost06 ?

Thanks in advanced



 Comments   
Comment by Cliff White (Inactive) [ 03/Mar/12 ]

You need to run tunefs.lustre --writeconf on the MDS also, have you done this? See Section 14.1.4 in the Lustre Manual, "Changing a Server NID"

Comment by Supporto Lustre Jnet2000 (Inactive) [ 04/Mar/12 ]

Yes, thanks. We follow this procedure:

1. unmount all the clients, and all the server.
2. tunefs.lustre --writeconf on the MDT and all the OST
3. on OST06 I do:
tunefs.lustre --erase-param --mgsnode=10.121.13.31@tcp --mgsnode=10.121.13.62@tcp --param="failover.node=10.121.13.59@tcp ost.quota_type=ug" --writeconf /dev/mapper/ost06p1
4 mount the mgs
5 mount the mdt and check with "lctl dl" that the osc are NOT present
6 mount all the ost
7 check with "lctl dl" that all the osc are present
7 mount the client

all was done...

thanks please close this issue.

Comment by Cliff White (Inactive) [ 06/Mar/12 ]

Please reopen if you have further issues relating to this.

Generated at Sat Feb 10 01:14:09 UTC 2024 using Jira 9.4.14#940014-sha1:734e6822bbf0d45eff9af51f82432957f73aa32c.