https://datahubproject.io logo
Join Slack
Powered by
Hey again there :slightly_smiling_face: I'm facing...
# all-things-deployment
r
Hey again there 🙂 I'm facing still some issues : while deploying on AKS clustter i did a deplloyment on a namespace "ns1" it is workign well, and now i'm trying the installation in a second namespace "ns2" ans i'm stuck : The the seconde prerequisites helm install, the ES masters shoudl be in running but ain't ! Still in pending, no logs on those pods , i don't get the point. If you have any idea by any chance 😕
✅ 1
s
Have you checked what is causing the Pending for elasticsearch? Are there enough resources?
Have you tried doing a describe for the elasticsearch deployment?
r
nop
Hi
just a kubectl logs elastic-master.-0
You may be on the good way
i cannot have 2 installs on that 3 nodes of DH
i
How many nodes does the cluster have?
r
3
the first install is great , healthy and stuff
but the second is failing as i said
how to check the possible issue as @square-activity-64562 mentionned
i
If you run 
kubectl get events --sort-by='.metadata.creationTimestamp' -A
do you see anything related to elasticsearch?
r
i
It is possible that some anti-affinity rule does not elastic to be deployed
There you go, your nodes are too small for elastic
r
or port already used or something
ok
i
The cluster needs more nodes or bigger nodes
Port conflict should not be a problem in k8s
r
yes i see
yep indeed
so i'll try adding new nodes
and see
2 or 3 more you think ?
And voilà ! with 2 more nodes, so we have now 5 nodes, i have ES master 0 ok, ES master 1 OK, and the last one still pending
Well the ressources were the problem ! Thanks @square-activity-64562 / @incalculable-ocean-74010
👍 1
i
One thing you can experiment with is sharing elasticsearch resources across DataHub instances.
Deploying all requisite software (elastic, relational DB & Kafka) in a namespace say: 
prereq
and then just deploy datahub components in another namespaces like 
dev
and 
prod
which connect to the same elasticsearch/db/Kafka instances in the 
prereq
namespace. You would lose isolation of stateful resources but its a better resource utilization altogether. Depends on your use-case :)
r
Yep i see but you can imagine one deployment / "customer" , each one must be isolated, including the prepreq.
i
Bare in mind that those stateful deployments are yours to manage. Our prerequisite charts do not offer any sort of maintenance or operational guarantees.
DataHub only components should be fine as they are stateless.
r
i think i'll use as mych as possible on cloud services
mysql DB mostly
elastic / kafka not really hosted solution
Open in Slack
PreviousNext
Powered by