Cache For OSD / Replication
pradeep suresh
32 Posts
November 30, 2022, 9:00 amQuote from pradeep suresh on November 30, 2022, 9:00 amWe had a couple of questions related to OSD Cache and replication,
We have 1.8TB NVMe SSDs on each of the 4 Petasan nodes and 600 GB capacity as cache is assigned for the OSDs on the nodes. What is a good cache size considering the capacity of the physical disks? If we add more physical disks to the nodes now, we may not be able to provision cache from the NVMe SSDs. Would it be ok to configure cache from non-NVMe SSDs for the OSDs?
What would happen if there is a hardware failure of cache and journal disk, what would happen to the OSD's associated to those disks?
Would be helpful if you could list the impacts.
Regradign the Replication :
When a replication task for iSCSI LUNs is setup between 2 Petasan clusters (with a frequency of 4 hrs), is the replication incremental where only changed fragments from the source cluster/LUN?
One of our use cases is to setup replication for a LUN and let it run for a few days (let’s say LUN A in source cluster to LUN B in destination cluster). Then we stop the replication and activate the LUN B at the destination and use it. Then again after a few days, we would want to shutdown LUN B and restart replication from LUN A. When replication is restarted from A to B, would that be a full replication or incremental changes ?
Thanks in advance
We had a couple of questions related to OSD Cache and replication,
We have 1.8TB NVMe SSDs on each of the 4 Petasan nodes and 600 GB capacity as cache is assigned for the OSDs on the nodes. What is a good cache size considering the capacity of the physical disks? If we add more physical disks to the nodes now, we may not be able to provision cache from the NVMe SSDs. Would it be ok to configure cache from non-NVMe SSDs for the OSDs?
What would happen if there is a hardware failure of cache and journal disk, what would happen to the OSD's associated to those disks?
Would be helpful if you could list the impacts.
Regradign the Replication :
When a replication task for iSCSI LUNs is setup between 2 Petasan clusters (with a frequency of 4 hrs), is the replication incremental where only changed fragments from the source cluster/LUN?
One of our use cases is to setup replication for a LUN and let it run for a few days (let’s say LUN A in source cluster to LUN B in destination cluster). Then we stop the replication and activate the LUN B at the destination and use it. Then again after a few days, we would want to shutdown LUN B and restart replication from LUN A. When replication is restarted from A to B, would that be a full replication or incremental changes ?
Thanks in advance
Cache For OSD / Replication
pradeep suresh
32 Posts
Quote from pradeep suresh on November 30, 2022, 9:00 amWe had a couple of questions related to OSD Cache and replication,
We have 1.8TB NVMe SSDs on each of the 4 Petasan nodes and 600 GB capacity as cache is assigned for the OSDs on the nodes. What is a good cache size considering the capacity of the physical disks? If we add more physical disks to the nodes now, we may not be able to provision cache from the NVMe SSDs. Would it be ok to configure cache from non-NVMe SSDs for the OSDs?
What would happen if there is a hardware failure of cache and journal disk, what would happen to the OSD's associated to those disks?
Would be helpful if you could list the impacts.Regradign the Replication :
When a replication task for iSCSI LUNs is setup between 2 Petasan clusters (with a frequency of 4 hrs), is the replication incremental where only changed fragments from the source cluster/LUN?
One of our use cases is to setup replication for a LUN and let it run for a few days (let’s say LUN A in source cluster to LUN B in destination cluster). Then we stop the replication and activate the LUN B at the destination and use it. Then again after a few days, we would want to shutdown LUN B and restart replication from LUN A. When replication is restarted from A to B, would that be a full replication or incremental changes ?
Thanks in advance
We had a couple of questions related to OSD Cache and replication,
We have 1.8TB NVMe SSDs on each of the 4 Petasan nodes and 600 GB capacity as cache is assigned for the OSDs on the nodes. What is a good cache size considering the capacity of the physical disks? If we add more physical disks to the nodes now, we may not be able to provision cache from the NVMe SSDs. Would it be ok to configure cache from non-NVMe SSDs for the OSDs?
What would happen if there is a hardware failure of cache and journal disk, what would happen to the OSD's associated to those disks?
Would be helpful if you could list the impacts.
Regradign the Replication :
When a replication task for iSCSI LUNs is setup between 2 Petasan clusters (with a frequency of 4 hrs), is the replication incremental where only changed fragments from the source cluster/LUN?
One of our use cases is to setup replication for a LUN and let it run for a few days (let’s say LUN A in source cluster to LUN B in destination cluster). Then we stop the replication and activate the LUN B at the destination and use it. Then again after a few days, we would want to shutdown LUN B and restart replication from LUN A. When replication is restarted from A to B, would that be a full replication or incremental changes ?
Thanks in advance