ForumGeneral Discussionceph journal partition

You need to log in to create posts and topics. Login · Register

ceph journal partition

hoanglongvina
12 Posts

August 31, 2017, 6:41 am

Hi Everybody,

Currently the default config of ceph journal when installing PETASAN is 5G.
I want up to size of the ceph journal partition? please help me, Thank.

admin
2,930 Posts

August 31, 2017, 9:30 am

You can change this value from the config file:

/etc/ceph/CLUSTER_NAME.conf

add this at end:

osd_journal_size = 10240

The value is in MB so this will make it 10G

you need to do this on all your osd nodes and restart the machine or restart the osd services

hoanglongvina
12 Posts

August 31, 2017, 10:21 am

Hi ADMIN,

Currently each partition has its own journal partition, I want to configure a SSD drive as its own journal partition ?

ex:

1. /dev/sdb : 20 G

/dev/sdb1: 15G --> DATA

/dev/sdb2: 5G --> Journal

2. /dev/sdc : 20 G

/dev/sdc1: 15G --> DATA

/dev/sdc2: 5G --> Journal

i wan't config 1 driver SSD:

---> 1 SSD 20 G --> Journal

Thank Everybody.

admin
2,930 Posts

August 31, 2017, 10:30 am

Hi there,

Currently we do not support mixing SSDs and spinning disks together as mentioned in our release notes, but it will be supported in the future when adding support for Bluestore. However you can do this now yourself manually using ceph-disk command.

hoanglongvina
12 Posts

August 31, 2017, 10:38 am

Hi ADMIN,

ex:

I'm create partition sdb:

/dev/sda : systems

/dev/sdb1 : 15 G ceph osd

/dev/sbb2: 5 G ceph Journal

when i add create partition journal + "5 G" = 10G , it's create "5G" to partition (sdb) ?

Thank.

admin
2,930 Posts

August 31, 2017, 11:03 am

Assume you have 2 spinning disks /dev/sdd and /dev/sde for OSD and SSD for journal on /dev/sdf

You can manually create the 2 OSDs with journal as follows

ceph-disk prepare --cluster CLUSTER_NAME /dev/sdd /dev/sdf
ceph-disk prepare --cluster CLUSTER_NAME /dev/sde /dev/sdf

This will create OSD partitions

/dev/sdd1 size is entire disk

/dev/sde1 size is entire disk

/dev/sdf1 SSD journal for first disk size determined by osd_journal_size in conf file (default 5G) as per earlier post

/dev/sdf2 SSD journal for second disk size determined by osd_journal_size in conf file (default 5G) as per earlier post

You do not need to create the partitions yourself, just specify the disk is fine

hoanglongvina
12 Posts

September 1, 2017, 3:01 am

Hi ADMIN,

I will try on the my systems. Thank very much.

hoanglongvina
12 Posts

September 1, 2017, 8:02 am

Hi ADMIN.

I'm setup petasan with partition OSD on /dev/sdc (driver - SAS)

/dev/sdc1 --> 10G Type Ceph OSD
/dev/sdc2 --> 5G Type Ceph Journal

I want dell or down partition ceph Journal sdc2 (5G --> 1M) and create partition ceph journal on one driver - SSD (/dev/sdd). mount it's with command

ceph-disk prepare --cluster CLUSTER_NAME /dev/sdc2 /dev/sdd --> can't create partition Ceph Journal.

please help me.

Thank.

admin
2,930 Posts

September 1, 2017, 9:48 am

The steps i sent earlier is for creating new OSDs with SSD journals.
If you have existing OSDs and need to move/replace their journals to SSDs, you can follow:

https://docs.hpcloud.com/hos-3.x/helion/operations/maintenance/ceph/replace_osd_journaldisk.html
https://www.sebastien-han.fr/blog/2014/11/27/ceph-recover-osds-after-ssd-journal-failure/

You may find it easier to use the steps i sent ealier for creating new OSDs. To stop PetaSAN from automatically creating OSDs itself during cluster creation or new node joining: during node deployment in the web deployment ui uncheck the "Local Storage Service" this will create a cluster (or join a new node) without creating OSDs then after the cluster is built (ignore warning that it does not have storage) go to the Node List and manually re-add/re-check the "Local Storage Service" role, now manually add your OSDs as per my prev steps.

hoanglongvina
12 Posts

September 4, 2017, 1:55 am

Hi ADMIN,