Ceph health Warn
tuxadmin
7 Posts
March 18, 2018, 2:06 pmQuote from tuxadmin on March 18, 2018, 2:06 pmHey !
Actually I built PetaSAN cluster with 3 osds for every node but it give me
ceph health warn
many no of pgs per osd .
kindly any advise to modify this .
how can i calc no of pgs per osd . there is recommended values .
also where i can modify it .
Thanks !
Hey !
Actually I built PetaSAN cluster with 3 osds for every node but it give me
ceph health warn
many no of pgs per osd .
kindly any advise to modify this .
how can i calc no of pgs per osd . there is recommended values .
also where i can modify it .
Thanks !
admin
2,930 Posts
March 18, 2018, 2:53 pmQuote from admin on March 18, 2018, 2:53 pmThe number of PGs should be determined at cluster/pool creation, It is not not a parameter that should be modified on a regular basis (Ceph allows you to increase the PG count via cli commands but not to decrease, also you can only increase it in steps and each will result in massive data movements). We compute it at cluster creation when the user selects the cluster disk size ( 5-15, 15-50 ..etc ). It is likely you have chosen a size range but not yet have added enough disks.
The only fix is to add more OSDs. it is not possible to shrink the current PG size. If you have a test install, you may re-install and select a smaller cluster disk count.
The ideal PG per disk is approx 100. Higher values will work but may put strain on your system, specially during periods of recovery and backfilling, hence the warning.
The number of PGs should be determined at cluster/pool creation, It is not not a parameter that should be modified on a regular basis (Ceph allows you to increase the PG count via cli commands but not to decrease, also you can only increase it in steps and each will result in massive data movements). We compute it at cluster creation when the user selects the cluster disk size ( 5-15, 15-50 ..etc ). It is likely you have chosen a size range but not yet have added enough disks.
The only fix is to add more OSDs. it is not possible to shrink the current PG size. If you have a test install, you may re-install and select a smaller cluster disk count.
The ideal PG per disk is approx 100. Higher values will work but may put strain on your system, specially during periods of recovery and backfilling, hence the warning.
Last edited on March 18, 2018, 2:56 pm by admin · #2
Ceph health Warn
tuxadmin
7 Posts
Quote from tuxadmin on March 18, 2018, 2:06 pmHey !
Actually I built PetaSAN cluster with 3 osds for every node but it give me
ceph health warn
many no of pgs per osd .
kindly any advise to modify this .
how can i calc no of pgs per osd . there is recommended values .
also where i can modify it .
Thanks !
Hey !
Actually I built PetaSAN cluster with 3 osds for every node but it give me
ceph health warn
many no of pgs per osd .
kindly any advise to modify this .
how can i calc no of pgs per osd . there is recommended values .
also where i can modify it .
Thanks !
admin
2,930 Posts
Quote from admin on March 18, 2018, 2:53 pmThe number of PGs should be determined at cluster/pool creation, It is not not a parameter that should be modified on a regular basis (Ceph allows you to increase the PG count via cli commands but not to decrease, also you can only increase it in steps and each will result in massive data movements). We compute it at cluster creation when the user selects the cluster disk size ( 5-15, 15-50 ..etc ). It is likely you have chosen a size range but not yet have added enough disks.
The only fix is to add more OSDs. it is not possible to shrink the current PG size. If you have a test install, you may re-install and select a smaller cluster disk count.
The ideal PG per disk is approx 100. Higher values will work but may put strain on your system, specially during periods of recovery and backfilling, hence the warning.
The number of PGs should be determined at cluster/pool creation, It is not not a parameter that should be modified on a regular basis (Ceph allows you to increase the PG count via cli commands but not to decrease, also you can only increase it in steps and each will result in massive data movements). We compute it at cluster creation when the user selects the cluster disk size ( 5-15, 15-50 ..etc ). It is likely you have chosen a size range but not yet have added enough disks.
The only fix is to add more OSDs. it is not possible to shrink the current PG size. If you have a test install, you may re-install and select a smaller cluster disk count.
The ideal PG per disk is approx 100. Higher values will work but may put strain on your system, specially during periods of recovery and backfilling, hence the warning.