Hi All,
I've been provided by this hardware:
4xHP server G8
18 HD 1TB per server (72 HD in total)
3 SSD per server (12 in total)
I do have now 2 questions:
1. How would it look like having 1xSSD performing journal each 6xOSD?
Is it feasible? What would you do?
2. I thought it would not be a very good idea to setup a cluster in
that way, and for this reason I setup the following crush map:
-1 69.96 root default
-14 64.8 region SAS
-12 32.4 datacenter 1-SAS
-3 16.2 host z-srv-m-cph01-sas
4 0.9 osd.4 up 1
....
-7 16.2 host z-srv-m-cph02-sas
84 0.9 osd.84 up 1
....
-13 32.4 datacenter 2-SAS
-8 16.2 host r-srv-m-cph01-sas
7 0.9 osd.7 up 1
....
-9 16.2 host r-srv-m-cph02-sas
5 0.9 osd.5 up 1
....
-15 5.16 region SSD
-10 2.58 datacenter 1-SSD
-2 1.29 host z-srv-m-cph01-ssd
1 0.43 osd.1 up 1
21 0.43 osd.21 up 1
45 0.43 osd.45 up 1
-4 1.29 host z-srv-m-cph02-ssd
83 0.43 osd.83 up 1
97 0.43 osd.97 up 1
90 0.43 osd.90 up 1
-11 2.58 datacenter 2-SSD
-6 1.29 host r-srv-m-cph01-ssd
3 0.43 osd.3 up 1
30 0.43 osd.30 up 1
62 0.43 osd.62 up 1
-5 1.29 host r-srv-m-cph02-ssd
2 0.43 osd.2 up 1
31 0.43 osd.31 up 1
61 0.43 osd.61 up 1
I would like to setup a replication in a way that:
pool_size 2 and 1 copy each Datacenter.
After I created 2 pools:
rbd-sas
rbd-ssd
And this crush_ruleset:
# ceph osd crush rule dump SSD
{ "rule_id": 1,
"rule_name": "SSD",
"ruleset": 1,
"type": 1,
"min_size": 1,
"max_size": 10,
"steps": [
{ "op": "take",
"item": -15,
"item_name": "SSD"},
{ "op": "chooseleaf_firstn",
"num": 0,
"type": "datacenter"},
{ "op": "emit"}]}
# ceph osd crush rule dump SAS
{ "rule_id": 2,
"rule_name": "SAS",
"ruleset": 2,
"type": 1,
"min_size": 1,
"max_size": 10,
"steps": [
{ "op": "take",
"item": -14,
"item_name": "SAS"},
{ "op": "chooseleaf_firstn",
"num": 0,
"type": "datacenter"},
{ "op": "emit"}]}
I associated the rbd-sas to SAS and rbd-ssd to SSD.
Do the 2 rule perform as I wanted? Sorry for the huge email.
Gian
_______________________________________________
ceph-users mailing list
ceph-users@xxxxxxxxxxxxxx
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
