Flushing is very slow


#1

Hello,

I recently installed couchbase 2.1.0 enterprise on Windows 2012 x64 and noticed that it takes FOREVER (eg. several minutes) to flush couchbase buckets either via the web ui or the .net cluster management api. To be clear, the number of items seems to have little or no effect (it it slow whether I have 10,000, 3 or even 0 items). Also, I am running as a single node only, if helpful.

Is there any reason why flushing operations would be taking so long?


#2

Replying with the hope it will subscribe to to further replies. I have noticed flushing is not the fastest thing in the world (kind of like rebalancing). It’s also usually impossible to set any items into the bucket while it’s flushing as well, which kind of makes sense.


#3

Hello,

In the current release it is true that the flush operation is quite slow, and it is a known issue that we are addressing.
it is recommend to take another approach, in this case to delete the bucket and recreate it.

Regards
Tug
@tgrall


#4

Is there a realistic timeframe when we can expect flush operations to be faster?


#5

Delete and recreate is similarly slow. The best recommendation at the moment is to create a bucket with a new name, and delete the old bucket.


#6

Deleting/recreating the bucket via couchbase-cli seems to be somewhat faster IF it works… However it is unreliable: in about 1 out of 5 of the attempts it fails:

ERROR: unable to bucket-delete; please check your username (-u) and password (-p); (500) Internal Server Error
{u’_’: u’Bucket deletion not yet complete, but will continue.\r\n’}

ERROR: unable to bucket-create; please check your username (-u) and password (-p); (503) Service Unavailable
{u’_’: u’Bucket with given name still exists’}

The blank line between the two errors was the output of bucket-list |grep <bucket_name>. So despite waiting till the ‘bucket-list’ no longer reports the bucket, creation fails.

[Couchbase-server enterprise edition 2.2.0]