Hard out of memory error with full ejection

Hello,
I’m using Couchbase 4.0 on 3 cluster machine, and i’m getting the hard out of memory error on one bucket.
The bucket is with full ejection.

In the vbucket 8MB on active user data in ram, 9MB on replica user data in ram and 4.71 GB on Total.
I putting the image bellow.

The bucket has 107M documents. We have an average of 400 op/s.

Any ideas why this is appening?

Thanks

Maybe its not memory but rather disk related… there is over 500 gigs of data there, could it be that this is reaching your disk limits? how much space do you have left on the hard drives?

Our cluster have 12TB of space.
Right now we are only using 1TB.

Hey @rafael.felix!

Looks to me like something else was using the memory here.

Were you building an index immediately prior to the out of memory error?

We have a view over that bucket, and because we have allot of operations the view is always indexing.
Could it be because of that?

It seems interesting to me because it looks as though all of your actual data has been evicted (which is why you have a resident ratio of 0).

Usually this is because memory is required for DCP, would you mind attaching some of the DCP graphs? You can find these in the UI

Sorry for the late response, but i was unable to come to the forum.
I didn’t take screen shot of dcp, and because i need to solve the ploblem i did a gracefull failover and restart the machines.

Next time i will take print screen of all UI to see if we can’t solve this problem.

Thanks

Hollo again,
I’m having the same error again. A bucket with full ejection is sending Hard out of memory.
Follows the images of dashboard.

I have 10T of free space.

Do anyone have an ideia why this is appening.

Thanks

The interesting statistic there is that there’s a total of 2.63GB of user data in RAM, even though the active and replica counts are under 1MB, so there’s certainly something not being accounted for.

I note you say you’re running 4.0 - is that 4.0.0 or some newer version? If you’re still on 4.0 I’d suggest updating to something much newer - 4.5.0 EE, or 4.1.0 CE. There’s been many issues fixed since the initial release of 4.0.

I’m using 4.0.0-4051 Community Edition.
I’m now updating to 4.1.0 Community Edition. Let me see if this problem comes again with that version.

Thanks

Hi there again,
I updated my cluster to the 4.1 CE and i’m having the error again.
Bucket is full ejection, i have 0% of active docs resident, on the vbuckets i have KB in ram, but in the total it says 5.45GB
I have free ram.
Can’t find out what is causing this.

Active and replica vbuckets have KB of user data but total have 5.4 GB.
Any more ideas?
Thanks

.

Could you please raise an MB (bug) on this: http://issues.couchbase.com. Please include the logs from the cluster: Couchbase SDKs

Do i need a login to raise the MB?
Can i use the forum login?

Hi,

Was this issue resolved? We’re having the same problem and considering to replace CB with another DB.