Increased memory not detected by Allocator

[Niklas (Niklas Bergström)]

Hi. We are trying ECE out and I installed it on a Redhat 7 with 16 GB. of memory.. We played around a bit, creating clusters e.g but 16GB does not fit that many clusters.
So we installed 32 GB in the server.
The allocator still shows a total capacity of 11.96 GB and I get "Unexpected error during step: [allocate-instances]: [no.found.constructor.steps.allocation.AllocationFailedException: Allocation failed: [NotEnoughCapacity(0,1,Map(zone-0 -> 2048))]]" when trying to start up an elastic with 2 GB on the allocator showing less than 1 GB free.

The free command on the OS shows 6.9 GB free memory.

Does anyone knows how to "refresh" the allocator to see the new memory size?

Running ECE 1.0.0

[Yuri (Yuri Tceretian)]

Hi Niklas,

I am not sure that we currently support allocator capacity changing.
However, you can use the following strategy:

1. Vacate allocator. See Move Nodes From Allocators
2. Delete runner. See Delete Runner
3. Remove all docker containers (ECE container's name starts with 'frc-') from the host
4. Install ECE node. See Install on Additional Hosts

By default, the installer allocates 85% of available RAM for Elasticsearch clusters, but you can override this value. See Installer Script Reference and argument --capacity in particular.

PS This will work if your ECE cluster has more than one node and there is enough capacity on other nodes to accommodate clusters from the node you want to upgrade.