datasources share a cluster

Hi,
I have a shared cluster, hosting two data sources, A and B, I watched that A is disabled automatically twice, so would two data source compete for historical nodes ?

I do not setup any tier group. Currently, total memory of historical nodes is large enough to hold both A and B.

Should I do some extra config to hold more than one data sources ?

Thanks

YL

Did you configure “druid.server.maxSize” correctly?

In general, there is no automatic disabling of datasource (unless drop rules are configured at the coordinator). when you say disabled, does that mean having value of “used” column be “false” in db or by disable you just mean that they are not loaded on the historical.

do you see any errors in the coordinator/historical logs?

– Himanshu

Yes

It’s 8G memory on each node, it’s a staging cluster, I set maxSize to be 4G, and heap 2G, 4 historical nodes, so the total capacity is 4G * 4G = 16G, right ?

There are two datasources, the total size is 5G + 9G = 14G, It’s expected that all segments can be held in memory, but I found one of the two is disabled by cluster, even after being enabled manually.

I am doing more insights on coordinator and historical logs

By default, segments are replicated twice on the historicals. That is, coordinator will assign one segment to 2 historical nodes. That means, if you have 14 G worth of segments then you need atleast 28G capacity on the historicals. You can configure the replication factor via load rules described at http://druid.io/docs/0.7.1.1/Rule-Configuration.html

– Himanshu