Membase Bucket Management - Adding and Deleting Buckets
Was wondering if there was any advice about the number of buckets a Membase cluster can have. Provided there is plenty of RAM to distribute across them is there any downside to have many buckets.
Another question is whether adding new buckets and deleting new buckets on a regular basis is a bad thing? I am thinking over potentially versioning buckets. So creating a new bucket that may replace an old bucket. Can anyone suggest any issues with doing this?
Thanks!
We are currently working towards a live deployment so these are just estimates, answers to questions you asked:
1 - we could start with 5 buckets but could end up with 10-15.
2 - we do plan to have replication enabled on all buckets across to start of with a dual node cluster.
3 - we are still trying to work this out, but are looking to initially have 2 8GB ram win 2008 servers making up the cluster and initially allocate the ram equally amongst buckets till we figure out which ones need more/less.
the idea is that we no more than once a week we create new buckets to replace old buckets and have them populated before deleting old buckets with old data in them. expecting that the old buckets service data while we do the creating and population of new buckets.
we will be using membase 1.7.0
Just to follow up here, that all sounds fine. We have some known issues with >20 buckets, but it's all on the management side and we're working on resolving them.
Perry
Thanks Perry, thanks for the pointer, will look out for an update on the >20bucket issue and make sure we don't grow above that for now.
>>Was wondering if there was any advice about the number of buckets a Membase cluster can have. Provided there is plenty of RAM to distribute across them is there any >>downside to have many buckets.
there is no downside in creating multiple buckets on membase. as a best practice you want to make sure you leave out some buffer between available_ram and how much memory buckets are using.
1- how many buckets you plan to use
2- replication factor
3- bucket sizes
4- RAM size on each node
5- how many nodes in the cluster.
also deleting and creating buckets on a regular basis is fine ( assuming you are using the latest 1.7.0 version) but it helps if you share more information about how often you intend to delete the buckets ?
e.g : deleting buckets with million items every hour or so when cluster is in production ? or as a maintenance plan once a week ?