Tweaking defaults for better perfs on higher load #29
Conversation
jotak
requested review from
mariomac,
OlivierCazade,
memodi and
jpinsonneau
|
with this kind of config (and also some tweaks on the CR side) I can safely run hey-ho like this: |
examples/kafka/topic.yaml
Outdated
| strimzi.io/cluster: "kafka-cluster" | ||
| spec: | ||
| partitions: 3 | ||
| partitions: 8 |
There was a problem hiding this comment.
I'd set something bigger, as long as a single FLP reader can handle multiple partitions with relatively small cost. E.g. 24 or 32.
jotak
force-pushed
the
tweak-defaults
branch
from
df638ea to
dfb6cb7
Compare
hack/loki-distributed.sh
Outdated
| echo "" | ||
| echo "Configure FlowCollector Loki with:" | ||
| echo " url: 'http://loki-distributed-distributor.netobserv.svc.cluster.local:3100/'" | ||
| echo " querierUrl: 'http://loki-distributed-query-frontend.netobserv.svc.cluster.local:3100/'" |
There was a problem hiding this comment.
I think, the namespace here should be network-observability
There was a problem hiding this comment.
the default namespace changed: https://github.com/netobserv/documents/pull/26/files#diff-d828b4887a8f430cebe46d5cc46bed8578de7639942a7e66bb6bfa5ad2b55220
There was a problem hiding this comment.
well, actually it should be ${NAMESPACE}
There was a problem hiding this comment.
Missed this changed, sorry.
|
@jotak - as discussed in scrum, could we also update Kafka storage to be using PVC instead of ephemeral? |
these seem to be better defaults to me (empirically tested)
sounds ok to you?