Depends, the hypervisor can be the one to manage IO caching for example, so your guest OS would be none-the-wiser.
It is however a tremendously bad idea in almost all cases… fsync exists for a good reason…
If things were THIS tight, having a 3-replicas setup with one of the replicas, handling...
It’s fine if you want to believe things that are wrong, but HBAs + storage systems like Ceph have long eclipsed hardware raid when it comes to top-tier storage solutions…
Write caches aren’t a magical feature of hardware RAID and can be done just fine with RAM + any JBOD based system…
Whatever...
That's a double-edged sword. If your RAID card messes up, it corrupts everything and you lost.
Software RAID is much more popular (even given unlimited budget, that is) these days, for good reason. Losing N drives is less likely than losing 1 card (so long as N > 1 anyway).
The problem with...
Hmmm I remember spending some time computing ideal log file sizes a while back, guess this is as good as any time to review it.
Running @Sim's snippet, I get the following:
On the current primary:
+----------+------------+
| per_min | per_hour |
+----------+------------+
| 8.01 MiB |...
I get your point, but if you were an XF dev, where exactly would you draw the line between "lean" and "just a demo don't do this"?
It is really difficult...
If the goal is "what is the minimal number of components needed to run Xenforo" then you would only have:
1. MySQL/MariaDB single node
2...
The fact that compose file example you got just above is already this long for a dev-only setup is exactly why no one shares one for production. Not because it's secret but because it's too specific to how you run things.
For example, you mentioned Elasticsearch, but in production you would...
Those are generally good answers, but to add on it:
1. Stay away from NFS if you can; it is the most easily available option but has rather poor reliability and limited POSIX compatibility
2. If you use redis, run replicas colocated with each XF instance; the difference between 0ms and 2ms...
ModSecurity is more or less destined to die as a rule engine.
The problem being that indeed its maintainer decided to drop it, and refused to do the necessary licensing changes such that OWASP’s CoreRuleSet team (who writes and maintains most of the rules people actually install) could maintain...
Not an expert with XF internals specifically, but we have many front proxies and many XF instances, and this all works without any issue in that regard.
My understanding is that sessions are stored either locally by default (which is indeed bad if you don't ensure stickiness of user:instance...
Host-local redis (ie 127.0.0.1) should generally be as fast as a socket (though sockets are technically superior, yes).
What is really important however is to have redis colocated with XF. Even in a low-latency private network (think ~1ms lat to redis), the difference between local and remote...
I didn’t think of doing that to be honest.
Mainly because I am not sure I would be very excited about it if I were them, since it means extra support tickets etc. if it gets traction and used by less technical customers not recognizing that it breaks some feature 😅
(And by that I mean that...
Ah that’s a good approach indeed; we do something similar in other places internally, so that shouldn’t be too hard to backport :)
if only Safari on iOS had adopted it before… sigh…
Hmm tbh I think it’s a fine tradeoff in that case. For the user it’d be transparent and not actually slow down anything (realistically the LP time should be higher that firing a non-awaited POST). It means less compute-saving on XF side though, but no more than currently so I would take that...
At the moment, for sure yeah. I’d certainly not expect it to work perfectly anyway.
However if we consider the fact that we have to regularly refresh a csrf token anyway especially for guests, that’s somewhat of an avenue that could be used to work around the guests user counts and thread views...