Simulator Discussions

vsim cDOT 8.3.2 delay between login prompt and full working system



  - vsim 8.3.2

  - on ESXi 6

  - w/4GB mem, 2-cores

  - single-node cluster


When booting up I usually watch ONTAP's console and wait for the point where the login prompt is issued to see that it's up and running. Recently I've noticed the ' whole cluster DB process' takes awhile longer to come. In particular I notice the cluster_mgmt interface doesn't respond. So I log into the node_mgmt interface and type 'net int show' and see only the node_mgmt LIF and some other commands don't work. CPU usage is high during this phase but after N minutes all the interfaces, vservers, etc become available. 


I'm wondering if other see this delay and if so, how long it is? I just saw one that took 30 minutes to come fully up. 



Yes, it is normal.


The prompt is available way before the cluster ring is synchronized and all the interfaces up on all the nodes.


Thanks for the feedback. I figured some delay was normal, but 30 minutes for a single-node cluster really makes me wonder if I've got something mis-configured. 


I rarely reboot these things (usually suspend) but I'm going to try and start keeping better records of reboot->cluster-up times.


PS: I use and see this on both macOS Fusion and ESX6. I routinely drop the VM size from 5gb or 8gb (on ESX) down to 3gb or 4gb (on ESX) and wonder if this is a price I'm paying for doing that.


30mn looks like a a stretch yes, but regarding memory, I wouldn't be surprised indeed, it's worth doing some tests with the original VM size.


I do remember seeing this many times, but I haven't hit it lately.  On first boot, the cluster gets through create and come up pretty quickly.  Then on subsequent boots after its been configured you get to the cluster shell a _long_ time before the cluster rings all come up.  I was just doing 'cluster ring show' until the DBs are online.  Theres another delay I used to hit during boot where it would hang with 1 second on the countdown for 5-10 minutes, but that seems to be fixed in newer loaders. 

If this post resolved your issue, help others by selecting ACCEPT AS SOLUTION or adding a KUDO.