ceph.conf [osd.X] and [mon.X] sections during import_cluster (wip-53)

classic Classic list List threaded Threaded
4 messages Options
Reply | Threaded
Open this post in threaded view
|

ceph.conf [osd.X] and [mon.X] sections during import_cluster (wip-53)

jcalcote
Hi Yaguang,

(NOTE: This topic relates specifically to the wip-53 branch.)

During our testing of import_cluster, we found that two buffers must be passed to VSM during cluster import:

1. A json-formatted crushmap (from ceph osd crush dump)
2. A ceph.conf file that MUST contain proper [osd.X] and [mon.X] entries.

Two questions

First, given that modern Ceph (version 10+) no longer requires the [osd.X] and [mon.X] sections to be in the ceph.conf file, is your team planning to remove/relax that requirement and use other mechanisms to determine the osd and monitor node information?

Second, does VSM ever examine the contents of /etc/ceph/ceph.conf after the initial cluster import or creation? (We're considering dynamically generating the required [osd.X] and [mon.X] sections and appending them to the actual contents of the ceph.conf file before passing it to the import_cluster method, but this will only work if VSM doesn't reexamine the ceph.conf file after import.)

Thanks in advance,
John
Reply | Threaded
Open this post in threaded view
|

RE: ceph.conf [osd.X] and [mon.X] sections during import_cluster (wip-53)

ywang19
Administrator

Hi John,

 

Yes, your comments are good supplement. For your first question, yes, we will retrieve osd.x and mon.x information through other way if it’s not provided in ceph.conf, but in alpha release, you’d feed them manually. For the second question, when you say ceph.conf, do you mean the one on ceph nodes? VSM doesn’t check the ceph.conf on ceph nodes after cluster is created, it will use a copy in the database. If you updated ceph.conf after cluster creation, you could use a tool called import_ceph_conf to update the one in database. But so far the ceph.conf update doesn’t trigger VSM to re-construct the cluster top.

 

 

-yaguang

 

 

From: jcalcote [via vsm-discuss] [mailto:ml-node+[hidden email]]
Sent: Saturday, December 05, 2015 12:54 PM
To: Wang, Yaguang
Subject: ceph.conf [osd.X] and [mon.X] sections during import_cluster (wip-53)

 

Hi Yaguang,

(NOTE: This topic relates specifically to the wip-53 branch.)

During our testing of import_cluster, we found that two buffers must be passed to VSM during cluster import:

1. A json-formatted crushmap (from ceph osd crush dump)
2. A ceph.conf file that MUST contain proper [osd.X] and [mon.X] entries.

Two questions

First, given that modern Ceph (version 10+) no longer requires the [osd.X] and [mon.X] sections to be in the ceph.conf file, is your team planning to remove/relax that requirement and use other mechanisms to determine the osd and monitor node information?

Second, does VSM ever examine the contents of /etc/ceph/ceph.conf after the initial cluster import or creation? (We're considering dynamically generating the required [osd.X] and [mon.X] sections and appending them to the actual contents of the ceph.conf file before passing it to the import_cluster method, but this will only work if VSM doesn't reexamine the ceph.conf file after import.)

Thanks in advance,
John


If you reply to this email, your message will be added to the discussion below:

http://vsm-discuss.33411.n7.nabble.com/ceph-conf-osd-X-and-mon-X-sections-during-import-cluster-wip-53-tp266.html

To start a new topic under vsm-discuss, email [hidden email]
To unsubscribe from vsm-discuss, click here.
NAML

Reply | Threaded
Open this post in threaded view
|

RE: ceph.conf [osd.X] and [mon.X] sections during import_cluster (wip-53)

jcalcote
Hi Yaguang,

Yes, I did mean the ceph.conf files on the storage nodes - specifically, /etc/ceph/ceph.conf. Thanks for the feedback - it's very helpful.

Another related question: What happens when the user modifies ceph settings in the VSM gui? I'm sure VSM updates the ceph.conf copy in its database, but does it not also need to update /etc/ceph/ceph.conf on the storage nodes - and probably send a SIGHUP to ceph to tell it to reload the configuration?

Thanks,
John
Reply | Threaded
Open this post in threaded view
|

RE: ceph.conf [osd.X] and [mon.X] sections during import_cluster (wip-53)

ywang19
Administrator

When the ceph.conf in database is updated, the updated file will also be distributed to all ceph nodes, and so far no logic to tell ceph to reload.

 

 

From: jcalcote [via vsm-discuss] [mailto:ml-node+[hidden email]]
Sent: Wednesday, December 09, 2015 5:21 AM
To: Wang, Yaguang
Subject: RE: ceph.conf [osd.X] and [mon.X] sections during import_cluster (wip-53)

 

Hi Yaguang,

Yes, I did mean the ceph.conf files on the storage nodes - specifically, /etc/ceph/ceph.conf. Thanks for the feedback - it's very helpful.

Another related question: What happens when the user modifies ceph settings in the VSM gui? I'm sure VSM updates the ceph.conf copy in its database, but does it not also need to update /etc/ceph/ceph.conf on the storage nodes - and probably send a SIGHUP to ceph to tell it to reload the configuration?

Thanks,
John


If you reply to this email, your message will be added to the discussion below:

http://vsm-discuss.33411.n7.nabble.com/ceph-conf-osd-X-and-mon-X-sections-during-import-cluster-wip-53-tp266p273.html

To start a new topic under vsm-discuss, email [hidden email]
To unsubscribe from vsm-discuss,
click here.
NAML