-
Type: Bug
-
Resolution: Duplicate
-
Priority: Major - P3
-
None
-
Affects Version/s: None
-
Component/s: Cloud Manager, Ops Manager
-
Labels:
-
0.1
Under Storage Engine, we currently say:
If you do not specify a storage engine, Cloud Manager uses MMAPv1 by default.
Cloud Manager's backup daemons use the default storage engine of the mongod unless told otherwise. For MongoDB 3.0, this is MMAPv1, but for 3.2+ it's WiredTiger. Also MongoDB 3.0 is past its end-of-life date, so we may want to update this part too?
Note that this is also incorrect in the same Ops Manager docs.
The MMAPv1 engine minimizes storage in the blockstore: all blocks are already compressed and padding is disabled. WiredTiger compression offers no further storage benefit.
Please note that Cloud Manager uses S3 blockstores, not MongoDB blockstores, so I believe these two sentences can be removed from the Cloud Manager docs.
Blockstores are still very common in Ops Manager. Even though we *strongly* recommend that no one run more than 1 mongod per server, customers still tend to co-locate their appdb nodes with their blockstore nodes. For this reason, we may want to mention in the Ops Manager docs that with WiredTiger they can tune the cache all the way down to fit more than 1 mongod per server, but they really can't do that safely with MMAPv1.