Hi,

My question is about the solrcloud cluster we are trying to have. We have a
collection with Tlog and pull type replicas. We intend to keep all the
tlogs on one node and use that for writing and pull replicas distributed on
the remaining nodes.

What we have noticed is that when the Tlog node goes down and let's say the
disk is also lost, when the node is brought back up (since it was the only
node holding Tlog replicas for all shards it is not deleted from state.json

1) There is no way to remove the node from the cluster since the remaining
pull cannot become leader

2) Since the disk is blank (like a new node), when solr comes up the Tlog
replicas remain down. If we create the core folders with only
core.properties the Tlog replicas become active without any data but in
this case pull replicas sync and become blank as well

Is there a way to stop this sync of pull replicas in solrcloud mode
temporarily while we populate the index on tlogs. We can do this in legacy
mode

We understand that we can have multiple tlogs as replicas to solve this but
wanted to know how we can stop this replication.

Thanks,
Sandeep

Reply via email to