Hi,
what are the major drawbacks when setting
action.write_consistency:one
We've had quite a few problems before when a few of our nodes were restarted and we though we'd be fine if we have a quorum of masters and at least 1 data node (we currently hold all data on each node, 3 nodes total, 2 replicas)
The documentation(here and there) only mentions:
To prevent writes from taking place on the "wrong" side of a network partition, by default, index operations only succeed if a quorum (>replicas/2+1) of active shards are available.
I feel like this is not a sufficient explanation...
A similar question regarding the risks got answered with:
Setting this to one will let you write data even if only one shard copy is available. What are the risks? If this single copy gets corrupted, you have data loss...
But isn't this the same as if I just had 1 replica?
To me the advantages seem to outweigh the disadvantages, since we do prefer high avalability and designed our architecture to withstand a loss of two nodes (or so we thought)
An updated documentation might be helpful.
This is also not apparent when reviewing the cluster state! It shows yellow after the primaries are there, but you are not able to tell why it is not indexing data.
Luca