-
Notifications
You must be signed in to change notification settings - Fork 654
Linux: zipdeploy API times out in synchronous mode #2776
Comments
It is by design that any request to App Service will time out after 230 seconds. Please use the async API instead. It lets you poll the API to know when it is done. Also, please try using the new Run-From-Zip, which should make your deployment much faster. See this announcement for details. Basically, in your Azure App Settings, set WEBSITE_RUN_FROM_ZIP to 1, and then deploy as normal using zipdeploy. |
Ah, 230 isn't documented anywhere I've seen but I think that's counter to the what is expected. I'll investigate using the async API as an alternative. Run-From-Zip sounds like exactly what we need, however we're using a Linux host and being experimental I'm nervous about relying on it for a production site. I think we're slowly being pushed to AWS for hosting, Linux and Node feel like second class citizens on the Azure platform. |
I've renamed the issue to better capture what it is about. |
@davidebbo: I believe I'm running into similar issues when using To be fair, the .war file is ~280 MB, so maybe that's a factor? I noticed that your Run-From-Zip announcement doesn't apply to Linux (yet). Do you have any ideas on how to avoid the 502 in the Linux ecosystem? |
@weltan I'm on the WIndows side, so I'll let the Linux experts comment further. @rramachand21, who can look at this? |
@rramachand21 should I open an issue on a separate repo, or is here fine? Didn't want to hijack the issue. |
hi @rramachand21, just following up here: any word on timeout issues related with the |
@rramachand21 @davidebbo |
@BenWalters I am no longer working on this project, so I will let @rramachand21 or others comment further. |
You guys could be suffering from this too. |
@phawxby Looks like it could resolve the issue. We aren't using ZipDeploy directly in the CLI, we've been using a ZipDeploy VSTS marketplace task. So think we'd need to implement the CLI tasks to get this running. Will have a look, thanks for pointing it out! |
Hi If the problem persists and is related to running it on Azure App Service, please open a support incident in Azure: This way we can better track and assist you on this case Thanks, Joaquin Vano |
Due to some pretty severe issues with VSTS deployments and performance we've using CircleCI to assemble our application, compress to a zip and deploy via the REST API. Unfortunately due to this we're having to deploy with all the node modules bundled inside and we've bypassed the npm install phase of the deployment script. This results in deploying a 50Mb zip file.
This all works, we have our application deploying and running. We're using the non-asynchronous API because we can use that detect when the application is deployed and due to fan-out out build pipeline isn't getting choked, the region deployments happen in parallel.
However the API lies, it's doing weird things instead of holding the connection until deployment is complete.
At a little over 4 minutes we've had this.
At a little over 4 minutes we've also had this
At 3 minutes we've had this.
Ultimately deployment ends up taking 10-20 minutes even on the premium tier Kudu hosts.
A few extra notes:
The text was updated successfully, but these errors were encountered: