We invite you to shape the future of IBM, including product roadmaps, by submitting ideas that matter to you the most. Here's how it works:
Post your ideas
Post ideas and requests to enhance a product or service. Take a look at ideas others have posted and upvote them if they matter to you,
Post an idea
Upvote ideas that matter most to you
Get feedback from the IBM team to refine your idea
Help IBM prioritize your ideas and requests
The IBM team may need your help to refine the ideas so they may ask for more information or feedback. The product management team will then decide if they can begin working on your idea. If they can start during the next development cycle, they will put the idea on the priority list. Each team at IBM works on a different schedule, where some ideas can be implemented right away, others may be placed on a different schedule.
Receive notification on the decision
Some ideas can be implemented at IBM, while others may not fit within the development plans for the product. In either case, the team will let you know as soon as possible. In some cases, we may be able to find alternatives for ideas which cannot be implemented in a reasonable time.
[EGO][System] - Improve cluster restart speed on large clusters with unavailable hosts
It took long time to restart the master in a large scale cluster, even there is no workload running. The problem is more significant for clusters with many scavenged hosts. An example is in a ~2000 scavenging hosts cluster, it could take >30min to restart EGO and see cluster become available, which is unacceptable for production cluster, and also a significant burden for maintenance team. The current investigations found the following points can be enhanced: 1) When restarts, the Master LIM checked all the compute hosts first, before trying to start EGO. We saw master LIM went through each dynamic hosts in the hostcache first, before trying to claim mastership and starting VEMKD. Because many hosts in the hostcache file can become unavailable, this operation could take really long, especially when there are many scavenging hosts.
We are thinking management hosts should be started first, before checking those low priority dynamic compute hosts.
2) Automatically remove unavailable hosts from hostcache file based on Resource Group The EGO_DYNAMIC_HOST_TIMEOUT can help us by automatically remove unavailable hosts from hostcache file, but we can't use it in the cluster that has both servers and scavenging desktops in it.
We only want to remove unavailable scavenging hosts, but not servers.
If servers are removed, a 100% utilization graph doesn't tell us that all servers are in good state.
What we need is to set this parameter at Resource Group level instead of Cluster level. For example, we only set this for Scavenging Resource Group, but not ComputeHosts or ManagementHosts RG.
3) Improve LIM polling performance for unavailable hosts So far we saw it always took 2-3 sec for master LIM to poll each unavailable host. Because we always have hundreds of scavenging hosts as unavailable, it always took long time to go through all the hosts. We want EGO to enhance this part to reduce the time spent on unavailable hosts.
Do not place IBM confidential, company confidential, or personal information into any field.