Update Redirect Loop

Are all containers using the exact same WP_REDIS_* configuration constants?

This is my full config block from my terraform chart. The WP_REDIS_PREFIX is the same as the database name every site connects to. This is a unique value for each site. The rest is the same for all other sites with a set of redis pods per node.

define( ‘WP_REDIS_HOST’, ‘redis-controller-master.redis’);
{{- with .Values.externalDatabase }}
define( ‘WP_REDIS_PREFIX’, ‘{{ .database }}’);
{{- end }}
define( ‘WP_REDIS_CLIENT’, ‘predis’ );

define( ‘WP_REDIS_SERVERS’, [
‘tcp://redis-controller-master.redis:6379?role=master’,
‘tcp://redis-controller-replicas.redis:6379?alias=replica-01’,
] );

This loop only happens when the core is updated to the next version of WordPress. The core install comes within the docker image used for the pods with the web server root folder shared between each pod of a site. The content for the site itself (wp_content) is stored on NFS volume, which is mounted into the public root folder. Let me know if I can provide any other information for this. The issue only seems to happen when an autoupdate within the active instance is triggered. If I run a build, and the full site is rebuilt, this issues does not seem to happen, but it is a bit difficult to tell. The issue does not happen every time the core is updated, and it is very sporadic.

If there are any useful pieces of information I can provide let me know. Next time I notice it, I am happy to run any diagnostics and share that before I fix the issue manually.

Have you tried a full cache flush before trying to update?

I have not done that, but I would think it would be possible to do that with some additional work on the cluster. However, for the time being I do have the issue at the moment, and I would be happy to provide any diagnostics on a couple sites before resolving it by rebuilding the individual nodes.

I’ve seen this issue reported 1-2 times a year, for almost a decade now and I’ve never been able to reproduce it or figure out what could cause it.

Unfortunately the only suggestion I have is to flush Redis when this loop occurs.