Support » Everything else WordPress » Need help on server cluster for WordPress

  • I need advice on how to configure a cluster for WordPress. I’m the website maintenance person, and I’m working with a network admin who doesn’t seem to be sure what to do.

    The cluster was set up when the blog we service suddenly saw a huge upswing in traffic. Here’s how the network guy describes it:

    The cluster was configured on the fly, when it was setup we literally needed it online in 2-3 hours.

    Physically we have 3 webservers behind a load balancer that sends traffic to the server serving the fewest clients.

    We do NOT serve the static content separately – this is something I had talked to the user about moving to but it was not something I really wanted to experiment on. I could do it from a webserver level but if we went with a separate static content webserver I would prefer if we did it off a subdomain that was also load balanced.

    Physically the dynamic files are being synced every 1 minute between servers. The advantage to this method is we do not have a single point of failure that a fileshare (like nfs) would create. Since we were initially dealing with very high load I did not think this was a good option. Even now I am not very comfortable with an NFS share simply because it would be a lot of load when something like a static subdomain would do what we want perfectly.

    Right now we have a single MySQL powering the entire cluster with no failover setup right now.

    The problem is that the WordPress Admin panel doesn’t work well for the user: whenever she tries to upload images or make changes to her theme, it bounces around between the three servers and often returns an error message because it can’t find one (“theme doesn’t exist” or “file doesn’t exist”).

    I’ve asked the network admin to re-config the cluster or cache so this won’t happen, but he doesn’t seem to know how. He suggests using FTP for file changes – which is completely unacceptable to the user, who just wants her WP Admin panel – or creating a separate subdomain for administration. The latter is not something I’ve ever seen, since the cluster-powered WordPress blogs I’ve worked on are already seamless.

    Other options he suggests are an NFS mount, which seems vulnerable, or a sticky session in the load balancer, which he doesn’t want to do because he says it’s inefficient.

    I don’t know what to tell him. The sites I maintain already have seamless clusters behind them. What I want is for the Wpadmin panel to work the way it’s supposed to for the user.

    What do we do?

    Thanks.

Viewing 8 replies - 1 through 8 (of 8 total)
  • whooami

    (@whooami)

    Member

    umm, we use clustering at work. Its pretty basic, how we do it. all servers have the same content.. and yes there is an NFS mount for uploads. Thats NOT insecure.

    we also, for some clients, have staging servers where the content is pushed out the live servers. the staging server use the same NFS mount. As long as theyre using private IPS, (we use 10’s) .. it shouldnt be an issue.

    This, btw, is less of a wordpress issue, and more of a administration issue. But realize that that content that is served from WP is all stored in the database..

    .. so file-specific 404s suggest something else, and dont jive (unless its specific to things that again, are user uploaded) — rsyncing between servers shouldnt be ncessary and wouldnt be necessary if there was an NFS mount for user uploads.

    lastly, and Im just curious, but unless these are your own collocated servers, why are you concerning yourself with NFS mounts being secure? Most people that handle this sort of things are worlds above the standard godaddy *technician*.

    Putting aside, that done properly, they arent insecure, he’s the networking/administration guy.

    Thanks for your reply.

    lastly, and Im just curious, but unless these are your own collocated servers, why are you concerning yourself with NFS mounts being secure? Most people that handle this sort of things are worlds above the standard godaddy *technician*.

    I’m not concerned; it’s the network guy who thinks NFS creates a single point of failure.

    whooami

    (@whooami)

    Member

    oh, well him he’s wrong then 🙂

    ask him if the fact that all those identical WP installs accessing the same shared mysql server does the same? 🙂

    NFS mounts — the actual mount point is owned by root, the files inside are owned by the user or apache, however it’s specifically set up. In other words, Apache doesnt own the symlink.

    It works for us.

    I’m confused as to how you can have anything other than a single point of reference for the uploaded files? It is a recipe for inconsistency to have multiple file shares serving the same app (ostensibly via the same “location” from the local point of view).

    If a single represented piece of storage is considered a “point of failure” then your chap needs to look at his storage layer and introduce redundancy into it through RAID arrays etc.. If that already exists (and I should imagine it does) then what’s the issue?

    <i>I’m confused as to how you can have anything other than a single point of reference for the uploaded files? It is a recipe for inconsistency to have multiple file shares serving the same app (ostensibly via the same “location” from the local point of view).</i>

    I agree. It seems like a basic thing to me. But the network dude has his cluster set up for load balancing with syncing every minute. He doesn’t like NFS and thinks it creates a single point of stress (the blog we’re supporting has a huge readership). I’m kind of flummoxed because obviously there are many WordPress blog with huge readerships, and they must be using something like NFS so that the WordPress admin panel will function correctly.

    What I would really like to be able to do is sketch out for this guy an ideal configuration for hosting a WordPress blog across multiple servers.

    whooami

    (@whooami)

    Member

    readership means nothing– that aside most heavily traffic’d sites are clustered and use exactly what he hemming and hawwing about.

    Yes, I meant traffic rather than readership.

    whooami

    (@whooami)

    Member

    I knew what you meant — readership/traffic has nothing to do with where images or user-uploaded content is served from, in the case of using an NFS mount.

    Apache sees a symlink’d directory the same way it would see any other directory. There is no distinction. There is NO extra security issue.

    A high traffic’d site is going to see a load regardless, and in fact, like you already attested, is going to serve up alot of 404s trying to find thaty content doing it the way its currently being done.

Viewing 8 replies - 1 through 8 (of 8 total)
  • The topic ‘Need help on server cluster for WordPress’ is closed to new replies.