Hey guys, I've been thinking about the idea of running a elasticsearch
cluster in AWS/EC2 mixing "normal" (on-demand or reserved) instances with
spot instances.
The motivation is simple: spot instances run just as fast as a normal
instance but with a much lower price/hour. With the "only" caveat being
that the instance can die at any moment (if the floating price goes higher
than your bid).
The idea is to have a 50/50 or 30/70 split between normal and spot
instances. For example, 4 normal instances and 8 spot instances.
My questions/doubts are:
- Has any one tried this kind of setup?
- One "safety" issue is that master/primary shards can be allocated in spot
instances which are prone to die. Anyway around this? Maybe allocate all
primaries in the "normal instances"? - The number of spot instances can't be much higher than the number of
normal instances. That would mean that, maybe, a shard woulnd't have a
replica in at least one of the normal instances, which could be risky.
Anyway around this?
Any thoughts?
Thanks
Felipe Hummel
--
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 elasticsearch+unsubscribe@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/elasticsearch/20674cb7-5359-477f-b5c0-b2ac9b08cfce%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.