Gateway node timed out, service didn't use other nodes in cluster

[edhans]

I had a gateway that hung over the weekend and about 20 reports failed to update. 

The message in the service for Refresh History was "Timeout occurred when sending or receiving a request."

The details were:

Cluster URI:	WABI-WEST-US-redirect.analysis.windows.net
Activity ID:	ccf178bc-c984-4433-a2d5-519274c8f31a
Request ID:	21a5635a-4109-4847-a8fb-7ded1d4a490b
Time:	2018-07-09 06:13:47Z

When I went to look at Settings, Manage Gateways, it said "You are good to go" but had a warning that one node was not responding.

I logged into that PC and launching the gateway app hung, so I rebooted. Then everything worked fine.

However, I had two other nodes that were operating just fine. Shouldn't the service have recognized that a node was timing out and tried another node? Isn't that the purpose of the cluster feature?

[v-yuezhe-msft]

@edhans,

Based on your description, you see "You are good to go" under Manage Gateways, which means your gateway is online and has no issues connecting to the data source. In this case, the request will not reute to other gateway nodes.

As per this article, if the primary gateway instance is not online, the request is routed to another gateway instance in the cluster.

Regards,
Lydia

[edhans]

That isn't what is happening though. It says "good to go" but it will say that even if there is a node offline. I think "good to go" means the cluster is working. 

However, the node the service was contacting was hung and timed out (my error message above) and instead of retrying another node, the report sync failed.

I think it is a bug in the logic of the service. If it times out, it should try anther node, not fail. 

[edhans · Accepted solution]

I've added it to uservoice, though I think this is a bug or design flaw, not a "new feature." 

https://ideas.powerbi.com/forums/265200-power-bi-ideas/suggestions/34818685-gateway-node-timeout-cau...