Semantic model refresh failing due to unreachable On-Prem gateway

[kims]

Report's semantic model (dataflow source) is failing with the following error.

Data source error: {"error":{"code":"DM_GWPipeline_Client_GatewayUnreachable","pbi.error":{"code":"DM_GWPipeline_Client_GatewayUnreachable","parameters":{},"details":[],"exceptionCulprit":1}}} Table: Sites Slicer.
Cluster URI: WABI-WEST-US-B-PRIMARY-redirect.analysis.windows.net
Activity ID: 76c4b8b6-ad1e-4ffa-9826-5b1a05eb6989
Request ID: 27d1167b-a1f8-022c-68df-139377acf6a7
Time: 2024-08-09 17:37:57Z

When it was first published, scheduled refresh worked without any issue.

But then the 'unreachable' error started popping up once every 2~3 days. I would then go to Service and manually refresh it and that would work. Dataflows would also just fail randomly with the same error.

I thought the error could be due to the outdated gateway, so I updated it yesterday and now I'm getting this error constantly and even the manual refresh doesn't work.

Table doesn't seem to be the issue because it changes with every error.

I've ran multiple Network ports test on the On-premises data gateway and the results are not consistent.

Sometimes all succeeds. Sometimes there are 3~5 fails.

Not only that, sometimes the test won't even run; doesn't generate the result and doesn't update the 'Last completed test' date time.

Could this be a firewall issue on the gateway machine?

2nd Question

This maybe a dumb question.

I used the dataflows as the source in Power BI desktop. Published it to the Service. That created the semantic model: this is the one that is failing to refresh.

So why does it look like the semantic model is using the gateway? Shouldn't it just pull data directly from the dataflows?

To put it into perspective;

Let's say there's a weather station, a weather channel, a weather report app.

This is the flow of data I expect:

Weather station -> weather channel -> weather app

But it feels like it's flowing this way?

Weather station -> weather channel

                           \_____________________>weather app

[SaiTejaTalasila]

Hi @kims ,

What is your data source?(Is it Oracle db).

Thanks,

Sai Teja 

[kims]

it's mysql 8

[kims]

Yes it is on VM, not sure about the actual configuration. I've reached out to our sysadmin before posting. Hopefully we can sort this out.

Thank you very much.

[v-yohua-msft]

Hi, @kims 

Have you solved your problem? If so, can you share your solution here and mark the correct answer as a standard answer to help other members find it faster? Thank you very much for your kind cooperation!

Best Regards

Yongkang Hua

[kims]

I'm still waiting to hear back from our system admin.

[lbendlin]

Is your gateway a single VM or a cluster of multiple VMs?  Your error pattern indicates the latter, and is often caused by one of the cluster members misbehaving.  The load balancer randomly picks the cluster member for the refresh and that's why the error only occurs sometimes.