We’ve been using Octopus Deploy (v.2.6.4.951) for a while now and even with retention policies in place, we see incredibly painfull upload times when uploading packages to the built-in repository even though we’re uploading from the server’s local disk. (it can take up to 3 minutes to upload a 2.5 MB package)
It ends up taking just over an hour to upload all the packages that make up our application, which is unacceptable.
Doing the same test on a local VM (much less ressources but empty or near empty repository), the uploads completes within a minute or so.
How can we improve that performance on our server?
Is there a number of packages in the repository after which the performance has been noted to greatly decrease?
Any advice you may provide would be greatly appreciated.
Thanks for reaching out. I’m gonna need some info about your environment to check what might be going on:
How many packages do you have on your repository?
Are your servers in the cloud?
Where you having this slowness on a previous version (like 2.5)? We did several improvements to the upload-download streaming on 2.6 and it should be faster
Have you tried resetting your Octopus Service? this rebuilds the indexes of the db, which might help with your situation
Could you share us one of your deployment logs for at least 2 different environments? We’d like to compare the speed between those 2.
Feel free to set this ticket as Private if you dont want your deployment logs to be public.
Thanks for getting back to me. Here are the answers you were looking for…
Close to 4000, we just recently decreased the retention policy (56 days to 17 days) and it’s taking forever deleting those who are older than the new retention rule.
No, Octopus is running on a physical server
Wasn’t as bad before 2.5 but, then again, we had less packages in the repository.
No, not yet. We wanted the new retention policy to finish deleting old packages first, but it’s been deleting packages for close to 2 days now… I think we might try this now.
I really don’t see what deployment logs have to do with anything… Deploying the packages is not an issue to any environment. It is rather fast. But, before we can deploy the packages, we upload them to the built-in repository and that’s where the problem lies. This is the part that is painfully slow…
You said the retention policy task was running for over 2 days. Has that process finished? If not, please cancel the task and then go ahead and reset the Octopus server
The deployment log would help me get a glimpse of how much time does it take to download a package from the repository.
Resetting the server seems to have done the trick. It now takes on average 20 seconds to push a package to the built-in repository instead of 3 minutes as we had seen previously…