we at AutoScout24 (http://www.autoscout24.de) have 3 independent clusters
in two data centers, fed with the same data and updated with around the
same frequency. The source of the data (an Oracle Database) resides in only
one of the data centers.
If you can afford this, you just don't have big network partitioning
problems. In case of one data center being isolated from the other, you
just have stale data in one data center, but no replication nor conflicts
nor data corruption issues.
We do not make clusters span over multiple locations and send traffic to
two of the three clusters, as one is just fed data and passively waiting
for traffic (hot standby). This later is reserved for catastrophic failures
in the main data center.
As we have independent indices which can handle the whole traffic, the data
originates from a database and building an index from scratch takes around
an hour, we do not back-up the index. We also automatically create a new
index with each new build and deployment process, all in a blue-green way
via Elasticsearch aliases.
I hope my description helps you decide on your architecture.
Am Dienstag, 20. Mai 2014 17:57:49 UTC+2 schrieb Yann Barraud:
Does anyone figured out how to get a working Elasticsearch installation
with 2 data centers in 2 different zones ?
How do you get consistency ? Is it an active/active configuration ? How do
you manage network outages ?
I heard geoclustering is not a good idea. So if I have an active/passive
installation, how can I prevent data loss ? And replicate indexes ? (Both
with 0.90 and 1.X versions)
Thanks for ideas.
You received this message because you are subscribed to the Google Groups "elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an email to email@example.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/elasticsearch/04e8b1c6-b77e-481c-b46e-00220e748ff3%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.