-
Type: Bug
-
Resolution: Done
-
Priority: Major - P3
-
None
-
Affects Version/s: 3.6.5
-
Component/s: Replication, WiredTiger
-
ALL
-
-
Storage NYC 2019-02-25
I just tried upgrading one of my RS from 3.4.15 to 3.6.5, thinking as 4.0 is now released 3.6 might be mature/stable enough now, but I had a very bad surprise, the CPU load increase about 4 times on the secondary (I didn't try the primary of course, I don't want to loose my cluster yet). As you can see on the chart bellow, my mongo usage is very stable and the secondary usually stays at 5% of total server CPU on 3.4.15, as soon as I upgrade to 3.6.5 it jumps to 15-25%. The load stayed exactly the same on the primary, and so does the CPU usage.
When I saw that I thought: maybe I should try re-syncing the secondary to benefit from the improvement of 3.6, so I did this, the first time it filled the disk during the initial sync and crashed mongo so I tried again and the second time it worked, but the load is exactly the same.
I also tried putting back 3.6.0 and 3.6.1 to see where the regression is, but both didn't boot for unknown reasons yet.
Is this expected? are you aware of any change that would cause this huge regression?
This is my mongo cloud manager account monitoring the RS if it helps: https://cloud.mongodb.com/v2/5012a0ac87d1d86fa8c22e64#metrics/replicaSet/5414cfabe4b0ce23e21b4b3b/overview
- depends on
-
SERVER-34938 Secondary slowdown or hang due to content pinned in cache by single oplog batch
- Closed
-
WT-3894 Timestamp queue implementation and statistics improvements
- Closed
- is related to
-
SERVER-36221 [3.6] Performance regression on small updates to large documents
- Closed
-
SERVER-37233 Increase in disk i/o for writes to replica set
- Closed