[SERVER-16860] Secondaries of same spec cannot keep up with high primary write load Created: 15/Jan/15 Updated: 03/Dec/21 Resolved: 19/Mar/15 |
|
| Status: | Closed |
| Project: | Core Server |
| Component/s: | Replication |
| Affects Version/s: | 2.6.7 |
| Fix Version/s: | None |
| Type: | Bug | Priority: | Major - P3 |
| Reporter: | John Page | Assignee: | John Page |
| Resolution: | Incomplete | Votes: | 0 |
| Labels: | None | ||
| Remaining Estimate: | Not Specified | ||
| Time Spent: | Not Specified | ||
| Original Estimate: | Not Specified | ||
| Operating System: | ALL |
| Participants: |
| Description |
|
If I put a high continuous write load onto a a replica set I fall behind on the oplog until I fall off the end. All three nodes are the same (High) spec - using write concern defaults and a multi threaded insert only workload ( OPLOG, Journal and Collection on same Drive). They are all interconnected by the same network, as is the client. I know I could use a higher write concern but this is a simple and likely test scenario and I would expect the replication mechanism to be as fast as any external loader. |
| Comments |
| Comment by John Page [ 15/Jan/15 ] |
|
Sorry 2.6.7 |
| Comment by Eric Milkie [ 15/Jan/15 ] |
|
john.page |
| Comment by John Page [ 15/Jan/15 ] |
|
3 AWS EBS Instances - type C4.2XL Custom Java loader pushing in records with 10 fields, _id index only, 32 threads with batches of 512, average size or record 2K Insert rate on Secondaries (x2) 8,000-9,000 per second. Once 800GB loaded - oplog too old and secondaries failed. Don't have any other logs as resynced secondaries manually. |
| Comment by Scott Hernandez (Inactive) [ 15/Jan/15 ] |
|
Please provide your reproduction, logs, stats, version information, configuration and any other information about your tests including the detailed results. |