Poll config on a 4 node Ceph

[yena]

Hello,
i'm testing a 4 node ceph ( 8 OSD ).
Each node have
2 HD Sata 4TB
2 SSH 150G ( dedicated to the ceph Journal ).

Is it correct this Pool config ?
Size 4
Min Size 3
Crush ruleset 0
pg_num: 200

( i have userd: http://ceph.com/pgcalc/ )

Probably i haven't understund the Size parameter beacuse the Max value on the web interf. is 3...
( I plan to survive to 1 node crash or 2 OSD Crash ... safe ? )
Thanks!

 

[dietmar]

I would use:

Size 3
Min Size 2

Replicating data 4 times is a bit too much ...

Also, ceph docs suggest to round the pg count to the nearest power of 2.

 

[yena]

I would use:

Size 3
Min Size 2

Replicating data 4 times is a bit too much ...

Also, ceph docs suggest to round the pg count to the nearest power of 2.

Ok
now i'm testing:

Size: 3
Min Size 2
pg count 512 (from calculator) .. but i calculate this: 8osd x 100 / 3 = 266

 

[nicolaiB]

pg count 512 (from calculator) .. but i calculate this: 8osd x 100 / 3 = 266

Also, ceph docs suggest to round the pg count to the nearest power of 2.

your calculated pg_count of 266 is greater than 2^8 (256), so the next possible pg_count is 2^9 (512).

 

[udo]

...
Each node have
...
2 SSH 150G ( dedicated to the ceph Journal ).

Hi,
I assume it's an typo and mean 150GB SSDs?!

150GB is an uncommon size (for me) - what kind of SSD is this?
Because not all SSDs are working fine as journal-ssd (speed and durability). This is the reason, why DC-grade SSDs are the standard for journal-SSDs.

Udo

 

[Q-wulf]

Here is an SSD for journals primer just in case (scroll to the bottom)
https://www.sebastien-han.fr/blog/2...-if-your-ssd-is-suitable-as-a-journal-device/