Unfortunately we don't have good observability at the moment on the redis instances themselves, so I can't easily give you a number/graph, but looking at the memory usage and the logs it seems it's pretty unfrequent (as in it seems not to have happened since the VM rebooted). Though I'm not fully confident on that statement.

Once I review the o11y of those instances we will have a bit more data, but we might not have historical one. So if it's not a lot of trouble, you can try to use it see if it's stable enough and take the decision then.

Keep in mind also that we use rbd only for persistance though, so if there's a reboot you might lose data too (it's meant to be a cache service in the end), that has not happened in 10 moths, but well, might happen at some point without notice.

Cheers

On 03/19 11:17, Roy Smith wrote:

I'm thinking about using the Toolforge Redis instance for a persistent data store. For the data I want to store, it's OK if there's occasional evictions due to memory exhaustion, but only for fairly small values of "occasional".

So, I'm curious what the situation is for this particular instance. How often does data tend to get evicted? If the answer is, "We haven't seen that happen in many months", then I'm fine. If the answer is, "It happens every day", then I need to be looking at a different data store. _______________________________________________ Cloud mailing list -- cloud@lists.wikimedia.org List information: https://lists.wikimedia.org/postorius/lists/cloud.lists.wikimedia.org/