1 Rookie
•
57 Posts
0
57
Does anyone know what kind of snapshotting the ME5 uses for replication? Clone, split-mirror?
We were recently considering the idea of replicating our ME5 to a ME4 that we have. After reading the docs, it looks like we would need to allow for 4 times the anticipated pool size (Source data, snapshot 1, snapshot 2, queued replication). We don't have enough capacity for that. Anyway, this led me down a rabbit hole in regard to how storage snapshots work. A description of how "clone" or "split-mirror" snapshots kind of makes me believe this is how the ME5 handles replication.
"Clone snapshots, also known as split-mirrors or full copy snapshots, reference all the data on a set of mirrored drives. This method involves duplicating the entire dataset, not only new or updated data, which can be used independently from the original."
Can anyone confirm?
DELL-Chris H
Moderator
Moderator
•
8.7K Posts
1
July 1st, 2024 18:40
The_LostIT_Guy.
So the spirit of your question seems to be getting at "Do we really NEED 4x the volume space if we are doing replication?" Is that correct, if not then let me know.
In talking about just the local snapshots used for replication, there will be multiple snapshots and each snapshot will take up some space. The way our replication snapshots work according to the admin guide is "When first created from the primary volume, the internal snapshots consume very little space but will grow as data is written to the volume. Just as with any virtual snapshot, the amount of disk space used by an internal snapshot depends on the difference in the number of shared and unique pages between itself and the volume. The snapshot will not exceed the amount of disk space used by the primary volume" -pg 46
The types you mentioned (clone/split mirror) do not grow over time based on changes but rather are like a whole extra copy of the data and would take up the same capacity as the volume initially instead of growing over time.
So this is where the caution, judgement, and good monitoring of your used capacity is needed if not allowing for the full recommended space: Even though our snaps take up very little space initially, they will grow depending on how much the data changes. Very often, this growth is under-estimated and results in the pool running out of storage space on its drives. There is always the potential though that somebody could completely overwrite all of the data on a volume since the time of the previous snapshot and if that is occuring regularly between all the snaps they will need that full recommended space.
"When you create a virtual pool, configure the size to allow for four times the anticipated size of the primary volume. This size will account for the primary volume, plus the same amount of space for each of the two internal snapshots, and possible queued replication. This is the maximum amount of space that you will need for replication." -pg 46
Also, if the suggested maximum space is not accommodated for, then in the right circumstance you can have a service impacting outage because of your SAN pool being completely out of space.
Let me know if this helps.
The_LostIT_Guy
1 Rookie
1 Rookie
•
57 Posts
0
July 1st, 2024 20:37
Haha. That is certainly how I originally ended up here, but part of it is genuine curiosity...my desire to know.
On page 44 under Initial Replication, it says "When a replication set is created, the primary volume and its internal snapshots all contain the same data." I took that as meaning the primary volume, S1 and S2 must also consume the same amount of storage space. This assumption is what led me to believe that the ME5 may use clone or split-mirror snapshots. I'm not sure if I just somehow missed what you had pointed out on page 46, or if I forgot it as quickly as I read it. Regardless, I see now that my understanding was wrong.
This was a fantastic response, Chris. I really appreciate the time and careful thought you took to write it up.
(edited)