In a discussion titled “Configuring a CDN - massive rebake coming up?”, ljpp is considering a global rebake of hundreds of thousands of posts due to CDN-related image display issues, possibly switching from CloudFlare to KeyCDN. zogstrip explains that the “low CPU” rebake method processes posts in batches of 500 every hour, which ljpp estimates could take a month — though they note it’s less critical for Amazon S3 URLs since bucket URLs still resolve, and CDN URLs can be gradually replaced. ljpp asks if the process can run unattended and whether it’s interrupted by updates or reboots, suggesting targeting only posts with S3 URLs. sam recommends running the rebake via console, especially for such a major change. sam also clarifies that if uploads are on S3, a dedicated S3 CDN must be configured for images to benefit — a step ljpp confirms they’re prepared to take, having tested it on a smaller community. Finally, ljpp reports rebaking at over 200,000 posts per hour with no noticeable site degradation, finding the process far more manageable than expected.
@zogstrip I may looking at a global rebake in the near future, and I have hundreds of thousands of posts. Unless CloudFlare fixes their stuff, I may be switching to KeyCDN or similar.
What does this low cpu method do, and how slow it actually is?
Okay, so we are looking at a month of throttled rebaking
Then again, that is not so critical for Amazon S3 URLs, since the bucket URL will resolve ok, and rebaking would quietly replace the URLs to CDN version over time.
Can one run this process in the background, start it and forget it?
Most likely it is interrupted by a Discourse update and/or reboot?
Maybe its better just to rebake the posts that match Amazon S3 URL…