This post is more than 5 years old
3 Posts
0
1689
May 28th, 2017 23:00
ECS 3.0 Multinode Initialization slow?
Hi All,
I have deployed a multi-node instance of ECS community edition with 3.0 code base.
There are 4 nodes, each with 4 x 120GB eager zero think provisioned LUNs, sitting on a SAN capable of almost 1000Mbyte/sec controller level throughput (from IOPS generation test utilities)
Deployment went smoothly, and I have created the storage pool across all 4 nodes.
However, the storage pool as still not finished initializing after 11 days.
When i check the node statistics, i see a 1Mbit/sec network activity and 5% CPU on all nodes doing whatever background tasks are needed.
The DT initialization stats got to 384 DT very quickly, then grew to 480 after about 6 days, and has been there ever since. I believe it needs to get to 1960?
Is there any way to connect inside the docker containers to change the priority of DTInit process?
There is no lack of CPU, network or storage resources in my vSphere cluster to support this.
I am waiting on this to complete (ie change to Ready status) so that i can create geo-replicated VDCs between this multi-node cluster, and a second one sitting in another network segment.
Any help would be appreciated!
Breckan
3 Posts
0
June 4th, 2017 16:00
Problem found !
In the second cluster, you need to create the Storage Pool, but DO NOT create a VDC.
Even though there isnt an explicit VDC entity, you can still retrieve the VDC key on the second cluster.
Then return to the primary cluster and create the VDC using the key and IP addresses as appropriate.
I am now seeing the replication group active, and data being consumed on the second cluster when data is only uploaded to the primary cluster - so working as per design.
Breckan
3 Posts
0
June 1st, 2017 21:00
Hi Adrian,
thanks for your reply. Whilst I agree that local Storage Pool setup seems to be OK without full readiness (eg Virtual Data Centre, Namespace and Buckets can be configured), the geo-replication feature doesn't work.
My understanding of the geo-repl setup process is:
- create remote storage pool
- create remote virtual data center
- create local storage pool
- create local virtual data center
- add remote VDC to local VDC list (which i assume will also include config info about the remote storage pool)
- create replication group (where drop down list should populate with local and remote VDC)
When I attempt to add the remote VDC name, remote VDC key, and remote VDC endpoint addresses, i get the following error:
"Virtual Data Center creation failure. Be sure that you have correctly entered the IP address(s) of online devices and that Data Services has completed initialization. Please correct input or retry after a few minutes. Service Error details: Error 7000 (http: 500): An error occurred in the API Service. An error occurred in the API service. Cause: error insertVdcInfo."
I have verified that all nodes of the local ECS cluster can ping the remote cluster (at least from the linux host VMs!).
I really want to be able to model a WAN service between the 2 clusters so that i can observe how the platform operates in a distributed deployment.....
Heiko
1 Rookie
•
6 Posts
0
November 1st, 2017 06:00
Hi Adrian,
unfortunately the wiki page doesn't exist any more?
Regards Heiko