Archive | informational RSS feed for this section

Updating WordPress with libssh (and what I did when it was broken)

By on June 14, 2009 in php, site meta, wordpress

This article was orig­i­nally about auto­matic updates over SSH2 not work­ing for me when I upgraded to Word­Press 2.8. Word­Press 2.9 is out now, and this prob­lem turned out not to be their fault.

After fil­ing a bug report and work­ing through it with the Word­Press team, a solu­tion was even­tu­ally found (see the final update to this post). If you're not inclined to skip to the end then here is a spoiler: turn off open_basedir or make the dec­la­ra­tion less restrictive.

If you're not expe­ri­enc­ing this prob­lem (and Google says you're land­ing here if you're look­ing for help set­ting up SFTP/SSH updat­ing for Word­Press), I don't think this write-up will be of much help. But feel free to stick around; I've got snacks!

Con­tinue Reading →

Comments { 0 }

Apache Admin Notes: Raw Performance Numbers

By on January 15, 2009 in apache, linux, site meta

In late Decem­ber I started look­ing at alter­na­tive web servers, hav­ing hit a point where I actu­ally asked aloud (though to no one in par­tic­u­lar) why I still use Apache for any http or php needs that may arise.

I sat down, installed nginx (nginx 0.6.33) from the EPEL (Extra Pack­ages for Enter­prise Linux) repos­i­to­ries and lighttpd (lighttpd 1.4.20) from the RPM­forge repos­i­to­ries (both third party repos­i­to­ries worth writ­ing about in detail at a later date). Apache was already installed (httpd 2.2.3), so a vir­tu­al­host was con­fig­ured to serve data from the same doc­root using all three servers, and PHP through fastcgi was con­fig­ured for lighttpd and nginx1.

I then set up a stock Word­Press 2.7 blog, some large images, some thumb­nails, and cre­ated what amount to approx­i­mately a 20k index page with noth­ing cached. Using ApacheBench to fetch 1000 requests, 100 requests at a time, I came up with some sim­ple num­bers show­ing how long it took to return the same page for each web server.

While I openly admit that the bench­marks pro­duced here are arbi­trary and super­fi­cial, they do give some rough insight (10,000ft overview level insight) as to how much faster nginx and fastcgi are over apache 2.2 and mod_php. lighttpd also per­formed admirably, and with less has­sle involved in set up than nginx, but I sus­pect that may just be pre­vi­ous famil­iar­ity with the configuration.

A table of num­bers (soon to be a graph, when I feel like mak­ing it) and some thoughts as to what they mean and why we (the web host­ing com­mu­nity) still use Apache reli­giously fol­lows after the jump.

Con­tinue Reading →

  1. Credit where credit is due. I couldn't have started to work out the nginx fastcgi con­fig­u­ra­tion with­out help from this post on redemp­tion in a blog. ↩

Comments { 0 }

Hosting control panels

By on June 16, 2008 in applications, site meta

For about 6 years I've pon­dered the nature of web host­ing con­trol pan­els. When I ran a web host, I played with CPanel, Plesk, and Direc­tAd­min. I've since dab­bled in the open source waters of Web­min and User­min, and of late I've even poked at VHCS and Open­Panel. I have, to date, not seen a sin­gle con­trol panel that is pre­cisely what I want (with Direc­tAd­min com­ing clos­est, and almost being a win­ner if not for its tiered reseller model aaaaannnd price tag). The real stick­ing points come down to the fact that I want:

  • The con­trol panel to write human-readable, rea­son­able con­fig­u­ra­tion files: This is where Web­min falls down on the job. The con­fig files are valid, but god are they awful to read.

  • No men­tion or hint of reseller any­thing: I just want admin­is­tra­tive accounts run­ning the show and hosted accounts with domains linked to them. Maybe some way for the hosted to mon­key with the non-essentials of their own accounts. So on the grounds of overkill, pretty much every com­mer­cial panel fails my sim­ple needs.

  • Email sup­port for some­thing that's not anti­quated: some pan­els use qmail (or some ver­sion therein, which hasn't had a proper update in years), some use send­mail (which is older than dirt and half as secure), and some use what­ever the fla­vor of the month is (read: what­ever ships on the sup­ported OS dis­tro). Me? I'm a post­fix guy. I'm not wed­ded to it per se, but it's updated, fea­ture packed, works well and its con­fig­u­ra­tion is only mod­estly arcane (qmail, by the way, fails due to hav­ing an arcane con­fig­u­ra­tion file struc­ture that doesn't plug into antivirus or spam very easily).

    Regard­ing IMAP and POP sup­port, I find that these pan­els often use their own "ques­tion­able" clients, or rely on courier. There isn't really any­thing wrong with that, I just like dove­cot more (to be fair, Direc­tAd­min sup­ports dove­cot last I checked).

    I don't want to even get started with autho­rized SMTP sup­port from these pack­ages. God, the hor­ror.

  • No byzan­tine depen­den­cies: Don't tell me that I need some god awful fos­silized encryp­tion library or ldap to make this sys­tem work.

    "Lis­ten here, Mr. Con­trol Panel, I don't use LDAP, and unless that is what you're using as a back­ing store for your host­ing infor­ma­tion, nei­ther should you."

To that end, I've begun play­ing with Ruby and Thor to whip up some sim­ple add/delete scripts (with ini­tial suc­cess!), and the biggest stum­bling block I've encoun­tered so far is just pars­ing out the accounts that exist.

The Apache con­fig­u­ra­tion file for­mat is… archaic in its own right, being nei­ther XML nor attribute: value pairs. It's read­able, but there is a rea­son that an entire perl pack­age (HTTPD::Config, which may or may not exist any­more) was pro­duced just to scrape httpd.conf files.

I fear that with­out step­ping back a lit­tle and engi­neer­ing a proper solu­tion I'm just going to wind up reim­ple­ment­ing what I encoun­tered when I did overnight sup­port for Real­ity Check Net­works (no link pro­vided, because, well, you know. Those dudes…).

Comments { 0 }