Post-migration troubleshooting: Gateway timeout when enabling plugins
Here's one that caught me out on a recent Drupal to WordPress migration. As is common with my projects, there were three parties involved: the client, an external development team and myself. The WordPress site was first built on the development team's server, after which it was migrated to my local environment. When everything was ready for beta testing, we moved the site over to the client's staging server on a newly activated account over at Kinsta. Eventually, we'd move it over to a live server, also hosted on Kinsta.
After some initial tests on staging, I found that deactivating and reactivating plugins would cause the site to hang and show a '504 Gateway Time-out' error. This happened when re-enabling some but not all plugins.
I initially suspected some misconfiguration at the hosting end because there were some hitches with the newly create account. At the outset, the account's server had an issue which Kinsta support needed to fix. For convenience, we'd also made use of Kinsta's free site migration service. This is where they'll migrate an existing WordPress site into their environment. Though it would have been easy enough for me to do, we thought to give it a try. In hindsight, this was a bit of a mistake. The site migration service itself was fine but it did end up causing some confusion. First, a miscommunication in the migration request caused them to create a temporary domain we didn't want. They helpfully solved this by giving us a second temporary domain. However, they'd also upgraded everything to PHP 7 in the process. All of these issues were possible suspects for the time-out error but turned out to be red-herrings.
It took some time to pinpoint the cause behind the Gateway Time-out error. I do have to say that Kinsta support were very responsive throughout the troubleshooting process. They eventually put a senior engineer on the case who found the problem. It turned out the problem wasn't to do with Kinsta at all. There was a leftover setting from the original development team's server. It was a valid format so didn't cause an issue on either my local server or my staging server. However, it apparently can cause issues with plugins and did on the Kinsta environment.
What was the setting? The WordPress upload path directory was set to the development team's server path e.g. /home/dev/public_html/sitename
. Throughout the migration, I'd been doing a database search-and-replace looking for their development domain. Somehow, as the site moved from different servers, that server path string remained in the database, only to cause a problem when the site landed in the destination server on Kinsta.
I'm not sure if there would have been any way to have caught this problem earlier. It's one of those obscure errors that are easy to overlook and take time to resolve. There's also no practical way to do a database search-and-replace for every imaginable string. I'll have to rack this one up to experience.