Hi,
I have a 5 machine cluster, triple replication, 1 SSD per machine. I have the following process class configuration:
machine 1: 2 storage, 2 stateless
machine 2: 2 storage, 2 stateless
machine 3: 2 storage, 2 stateless
machine 4: 2 storage, 1 proxy, 1 stateless
machine 5: 1 log, 3 stateless
I’d made this configuration after referring to https://forums.foundationdb.org/t/questions-about-process-classes-recruitment-behavior-and-cluster-configuration/1177 . But I keep noticing my storage classes also being assigned log roles, automatically, leading to a warning that it’s not ideal to have the same process be both a storage and a log process.
How do I avoid this from happening? Any ideas where I’m going wrong?
gaurav
(gaurav)
June 8, 2020, 1:19pm
2
How many tlogs are there in the cluster - probably 3+ due to triple replication ? There appears to be only a single fdb process with log class; so if there are more tlogs then they will go to other fdbservers. From what you’ve configured, storage
seems like the next best alternate class for them.
See this post for more details:
Hi All, based on suggestions from a github issue, I am posting a visual representation of class/roles fitness matrix, and fitness ordering matrix:
[57%20PM]
@alexmiller @ajbeamon going through this matrix again, a few questions came up that I had probably not thought about earlier: How does FDB decide how many SS to recruit in the cluster? Will it try to recruit one SS on each process class where it is allowed to do so (that is everything except for those with NeverFit)?
Specifically, will …
1 Like
Ah, rookie mistake. This makes sense. Useful post too, will check it out.