Replication implementation state
Hi, I've been reading through the docs and replication is mentioned multiple times but always marked as "not available in the current version".
I actually noticed that on github the libvbucket already handles the number of replicas parameter.
I'm wondering what's the current state of replica implementation and if it's somehow possible to have it already running in some form.
I would also like to know if there's any way to estimate the amount of data loss (data in RAM but not persisted on disk yet) in case of a single node failure.
Thanks
Paolo, to follow up: When looking at the output of 'stats' to the server you will see these two values:
STAT ep_queue_size 0
STAT ep_flusher_todo 0
The sum of those would equal the amount of data that has not yet been persisted to disk.
Hope that helps, let me know if you have any other questions.
Thanks, take care.
Perry
Hi Paolo, thanks for the inquiry.
I believe the docs are referring to the ability to change the replica count. In the current beta (beta 2) the replica count is hard coded to 1 (meaning one "replica", 2 copies of any piece of data). In beta 3 (to be released very soon) this will become a configuration option.
In terms of unpersisted data, I will check to see what our available statistics are and get back to you.
Perry
Forum support is great for free but sometimes you need a guaranteed response time and dedicated resources for your questions or issues.
Consider purchasing enterprise-level support from Membase: http://www.membase.com/products-and-services/overview
Call or email "sales -at- membase -dot- com" today!