[horde] Issue when trying to use admin interface in recent update

[Nels Lindquist]

On 2026-05-12 9:19 AM, Nels Lindquist wrote:
> On 2026-05-12 1:57 AM, Ralf Lang wrote:
> 
>> MGET only works if all keys are available from one node. In your setup 
>> this might not be the case. I have released a fixed version of the 
>> library which will fall back to GET if MGET fails AND LOGS.
>> The new library also provides a method to suppress MGET when the admin 
>> knows it won't work but currently there is no horde config option to 
>> support this yet.
> 
> Interesting... it *shouldn't* be the case; I checked both nodes, and 
> neither of them have "cluster-enabled yes" or any other cluster 
> directives, so there should be no auto-sharding happening. All of the 
> keys should be available from whichever node is designated as master by 
> the sentinels.
> 
> Is there a way for me to test that functionality directly? I could also 
> switch to a newer redis version pretty easily if that might help--both 5 
> and 6 are available as distro modules or I could use Remi's repository 
> for 6.2 and higher.

Some additional info:

I updated the hashtable parameters in conf.php to use only the currently 
active redis host on port 6379 and set replication to 'none'.  The admin 
stuff immediately started working with no issues--I didn't even need to 
refresh my login. It feels like the new library might not be handling 
the sentinel setup correctly?

-- 
Nels Lindquist
nlindq at maei.ca