[SERVER-5423] oplog length affect on memory Created: 27/Mar/12  Updated: 15/Aug/12  Resolved: 27/Mar/12

Status: Closed
Project: Core Server
Component/s: Performance
Affects Version/s: 2.0.3
Fix Version/s: None

Type: Question Priority: Major - P3
Reporter: rajesh bhatt Assignee: Andy Schwerin
Resolution: Done Votes: 0
Labels: performance, replicaset
Remaining Estimate: Not Specified
Time Spent: Not Specified
Original Estimate: Not Specified
Environment:

RHEL 5.6
one primary and three standby replica sets.
Local disk storage with RAID 10
RAM : 72GB
Db size: 73GB
OPLOG : 20GB


Participants:

 Description   

Hi,

We want to change the oplog size to 50GB on primary database because of the stale replication issues.
after increasing the oplog,Will the local database(50GB) would be a part of mapped memory by mongod?

what should be the impact on overall database performance in a replication/journal enabled environment, if the size of the database is larger than the available physical memory on the system.

Thanks,
Rajesh Bhatt



 Comments   
Comment by rajesh bhatt [ 27/Mar/12 ]

Thanks Andy,

You have provided the exact information that I was looking for. For future I will be using the google group. We have the same environment as you have mentioned in the exceptions. Also we are looking into the sharding as well.

Comment by Andy Schwerin [ 27/Mar/12 ]

For future support requests, please try the mongodb-user google group, https://groups.google.com/forum/?fromgroups#!forum/mongodb-user or commercial support.

Comment by Andy Schwerin [ 27/Mar/12 ]

The whole oplog may be mapped into mongod's virtual address space. However, the portion of the oplog in the working set of mongod will depend on the rate at which data is written by clients. If your write load is so high that the secondaries get nearly 50 gigabytes behind the master, you're going to see an adverse performance impact.

You should consider sharding if you're having trouble sustaining your desired write load. If your secondaries are getting over 20 GB behind the master, I wouldn't expect growing to 50 GB is going to fix your problem. The exception to this is if your writes come in bursts, followed by sustained periods of little, or no, write activity.

Generated at Thu Feb 08 03:08:51 UTC 2024 using Jira 9.7.1#970001-sha1:2222b88b221c4928ef0de3161136cc90c8356a66.