We observe strange behavior with some configurations. PGs stays in degraded
state after
a single OSD failure.
I can also show the behavior using crushtool with the following map:
----------crush map---------
# begin crush map
tunable choose_local_tries 0
tunable choose_local_fallback_tries 0
tunable choose_total_tries 50
tunable chooseleaf_descend_once 1
# devices
device 0 osd.0
device 1 osd.1
device 2 osd.2
device 3 osd.3
device 4 osd.4
device 5 osd.5
device 6 osd.6
device 7 osd.7
device 8 osd.8
device 9 osd.9
device 10 osd.10
device 11 osd.11
# types
type 0 osd
type 1 host
type 2 rack
type 3 row
type 4 room
type 5 datacenter
type 6 root
# buckets
host prox-ceph-1 {
id -2 # do not change unnecessarily
alg straw
hash 0 # rjenkins1
item osd.0 weight 1
item osd.1 weight 1
item osd.2 weight 1
item osd.3 weight 1
}
host prox-ceph-2 {
id -3 # do not change unnecessarily
# weight 7.260
alg straw
hash 0 # rjenkins1
item osd.4 weight 1
item osd.5 weight 1
item osd.6 weight 1
item osd.7 weight 1
}
host prox-ceph-3 {
id -4 # do not change unnecessarily
alg straw
hash 0 # rjenkins1
item osd.8 weight 1
item osd.9 weight 1
item osd.10 weight 1
item osd.11 weight 1
}
root default {
id -1 # do not change unnecessarily
# weight 21.780
alg straw
hash 0 # rjenkins1
item prox-ceph-1 weight 4
item prox-ceph-2 weight 4
item prox-ceph-3 weight 4
}
# rules
rule data {
ruleset 0
type replicated
min_size 1
max_size 10
step take default
step chooseleaf firstn 0 type host
step emit
}
# end crush map
--------------------------------------
After compiling that map to ''crush-test.map' we run:
# crushtool --test -i 'crush-test.map' --rule 0 --num-rep 3 --weight 11 0
--show-statistics
I set '--weight 11 0' to mark osd.11 as 'out'. The result is:
...
rule 0 (data) num_rep 3 result size == 2: 111/1024
rule 0 (data) num_rep 3 result size == 3: 913/1024
so 111 PG end up in degraded state. I would expect that the data gets
re-distributed to the remaining OSDs instead.
Can someone explain why that happens?
_______________________________________________
ceph-users mailing list
[email protected]
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
