Forum Discussion

h_l_de_boer's avatar
h_l_de_boer
Frequent Visitor
2 days ago
Solved

Fabric Virtual network data gateways - offline

We have setup a Virtual network data gateway to connect to some Azure resources. This worked fine; we can connect to sql server in Azure, or to a Storage Account. But after some time, the Virtual ne...
  • jubinsoni's avatar
    2 days ago

    Hi h_l_de_boer​,

    Besides the known certificate issue already linked, it's worth ruling out that the gateway is simply asleep. That is by design, and it is the most common reason a VNet data gateway "works for a while, then shows offline".

    A VNet data gateway is a managed container that is deallocated after a period with no queries. In the UI that shows up as offline/inactive. It starts again on the next request, but that cold start can take a few minutes, and a refresh or a test connection that arrives in that window can fail.

    Manage connections and gateways -> Virtual network data gateways
    -> select the gateway -> Settings
    -> "Inactivity time before sleep" (default is short)
    -> raise it to cover the gap between your scheduled refreshes

    If raising that doesn't explain it, these are the other things I would check, in order:

    capacity    -> F SKU paused / scaled down overnight? gateway goes with it
                  (common when the capacity is paused by automation to save cost)
    subnet      -> delegated to Microsoft.PowerPlatform/vnetaccessLinks,
                   enough free IPs for the number of gateway members
    DNS / NSG   -> private endpoints resolve from that VNet;
                   no NSG / route table change blocking outbound
    certificate -> private CA SQL cert (known issue above)

    A practical workaround while testing: schedule a lightweight refresh, or a pipeline Lookup on a tiny query, just before the main refresh window. The gateway is then already warm when the real load starts.

    So my overall approach would be:

    check gateway status right before a failure -> asleep vs truly down
    asleep -> raise inactivity time before sleep / warm it up first
    down   -> capacity paused? subnet IPs / delegation? DNS + NSG?
    cert   -> register root CA (known issue mitigation)

    AI-assisted drafting: AI was used to help structure and phrase this response. I reviewed and validated the technical content before posting.