Bluehost have been a pain, but it is all over now.

Written By: - Date published: 4:59 am, June 30th, 2009 - 34 comments
Categories: admin, crosby textor, scoundrels - Tags: ,

Sysop

Sysop

Ok. The new server is up and running after a few painful days with the system being suspended, and then after moving it to a backup and having a partition run out of space (yesterdays problem).

The ostensible reason that Bluehost were suspending my account was because the processing usage on their CPU’s was exceeding their (unspecified) limits.  Bluehost thought that suspending the account manually and sending me a e-mail after they’d done it was an appropriate way of handling the issue. The way that this was expressed to me was that there was likely to be a rogue process running on my account.

I wasted considerable time looking for problems in the code and processes. This included staying up most of the night on Friday morning because all they could show me were some evidence of zombie processes from the day before. So I started looking at the server processes at the time that the issue was apparently happening the previous night. Eventually I went to bed. Half an hour later the dipshits suspended the account again. The ‘evidence’ was from when I was watching the site. The only CPU usage was from spiders, crawlers and other ‘bots scanning the posts and comments for the day.

My other thought had been that the site was being targeted in a denial of service attack. However there is no evidence of that either. A careful run through the logs shows that there was a substantial increase in traffic going on. Over the last 5 weeks, there has been close to a third increase in traffic on the site and not much from the ‘bots. The rise in traffic is largely in page-views of the front page and an increase in RSS feeds. Looks like the Mt Albert by-election, Worth, super-city, and other issues have been ramping interest up in political commentary. Hardly surprising as this government seem to be descending into a defensive cycle of  governing incontinence.

About the time I’d figured all of this out, the arseholes at Bluehost had cut me off in their usual version of ‘good’ service. I’d come to the conclusion that Bluehost simply didn’t like sites with growing traffic. Bluehost had been fine for the last 5 months while the site grew back from the post-election and holiday drop in traffic. Now it is up to election month levels, which is obviously more than they care for. Mind you, I can understand why.

Over the weekend I shifted the site to a backup server. It was slow and unfortunately a bit short of disk space. The latter I found out after I went to work on Monday and the database went down while I was reading the mobile version on the bus. Turns out that the log files  grow pretty fast these days and while there was a lot of room across the hard disks, there was very little on the disk partition running the database. But what was interesting was that even on the weekend the site traffic was showing interesting graphs of CPU usage and traffic – like averaging 30% CPU on the relatively quiet weekend traffic. I did some optimisation which should help on the new site, but I’ve got to seriously look at how to channel the site growth so it stops blowing out our servers.

Now we’re back on a server far from Crosby Textor and their habit of using litigation to hide their nasty political tactics. Fools like that, along with the costs, was the primary reason for shifting offshore. It makes it harder to use money as a gagging tool.

This time the server is with a technically orientated company rather than one looking at cost-cutting. Hopefully we’ll now get another 12 months before the site growth forces another crisis.

My apologies for the break in service. But The Standard is a site run on the smell of an oily rag with a lot of volunteers who have jobs and lives outside of politics. And on that note, I’d better hit the sack so I can code in a few hours.

G’night

34 comments on “Bluehost have been a pain, but it is all over now. ”

  • Comments are now closed

  1. infused 1
    30 June 2009 at 6:05 am

    Overseas sites are slow zzzz. Should have put the word out for some discounted hosting, im sure many would have helped rather than shifting it overseas.

  2. infused 2
    30 June 2009 at 9:02 am

    btw, site is really slow.

  3. infused 3
    30 June 2009 at 9:15 am

    some info for you. From home, cable connection. Takes 15-20 seconds for a page load.

    Tracing route to thestandard.org.nz [69.55.236.241]
    over a maximum of 30 hops:

    1 <1 ms <1 ms <1 ms 192.168.3.1
    2 * 12 ms 14 ms ****
    3 11 ms 11 ms 11 ms ge0-0.v4wlg0.acsdata.co.nz [202.21.136.193]
    4 21 ms 19 ms 21 ms ge0-0-35.v4akl1.acsdata.co.nz [202.61.3.2]
    5 20 ms 20 ms 19 ms Gi1-0.415.gw1.akl1.asianetcom.net [202.147.41.2
    5]
    6 177 ms 175 ms 174 ms po8-1.gw1.sjc1.asianetcom.net [203.192.185.50]
    7 146 ms 149 ms 246 ms ip-202-147-50-250.asianetcom.net [202.147.50.25
    ]
    8 * * * Request timed out.
    9 185 ms 188 ms 188 ms 72.165.184.6
    10 159 ms 160 ms 160 ms ge1-2-6509-a.castleaccess.com [69.43.169.112]
    11 160 ms 161 ms 161 ms 69.43.129.84
    12 160 ms 159 ms 160 ms 69-55-233-164.in-addr.arpa.johncompanies.com [6
    .55.233.164]
    13 160 ms 162 ms 159 ms 69-55-233-153.in-addr.arpa.johncompanies.com [6
    .55.233.153]
    14 159 ms 160 ms 161 ms virt12.johncompanies.com [69.55.227.70]
    15 * * * Request timed out.
    16 * * * Request timed out.
    17 * * * Request timed out.
    18 * * * Request timed out.

    From work citylink

    1 <1 ms <1 ms <1 ms *****
    2 1 ms 1 ms 1 ms ******
    3 2 ms 2 ms 2 ms dts-slx-gw.dts.net.nz [202.68.80.1]
    4 4 ms 4 ms 3 ms 172.24.5.97
    5 * * * Request timed out.
    6 12 ms 13 ms 13 ms 172.24.5.105
    7 12 ms 15 ms 11 ms Gi1-2.gw1.akl1.asianetcom.net [203.192.167.173]

    8 138 ms 139 ms 141 ms po8-1.gw1.sjc1.asianetcom.net [203.192.185.50]
    9 166 ms 166 ms 166 ms ip-202-147-50-250.asianetcom.net [202.147.50.250

    10 * * * Request timed out.
    11 299 ms 308 ms 299 ms 72.165.184.6
    12 153 ms 153 ms 155 ms ge1-1-6509-a.castleaccess.com [69.43.169.80]
    13 154 ms 152 ms 152 ms 69.43.129.84
    14 152 ms 153 ms 153 ms 69-55-233-164.in-addr.arpa.johncompanies.com [69
    55.233.164]
    15 183 ms 156 ms 152 ms 69-55-233-153.in-addr.arpa.johncompanies.com [69
    55.233.153]
    16 151 ms 152 ms 154 ms virt12.johncompanies.com [69.55.227.70]
    17 * * * Request timed out.
    18 * * * Request timed out.
    19 ^C

  4. lprent 4
    30 June 2009 at 9:17 am

    Yeah I’m puzzled about that. I can’t see any obvious slowdowns.

    But it is definitely slower. I just pulled the YesVote as being the one thing that isn’t cached

  5. infused 5
    30 June 2009 at 9:19 am

    Here is the load times… 😉

    http://iforce.co.nz/i/r3b3j2jg.jpg

    very, very, very slow… Cheaper isn’t better.

  6. Maynard J 6
    30 June 2009 at 9:28 am

    4:59 am? To classify as ‘a few hours’ sleep you need more than two – I am not sure you would have got that. Thanks, as always, for the work you put into this site. Category seems very appropriate fror Bluehost.

    • lprent 6.1
      30 June 2009 at 3:44 pm

      Lyn dragged me out of bed after I had about 3 and a bit hours. I usually only sleep for 5 or 6 hours anyway. However this site is seriously cutting into the reading I promised myself after the by-election

  7. Andrew 7
    30 June 2009 at 9:54 am

    Site seems to be slower than it has ever been 🙁

  8. jarbury 8
    30 June 2009 at 10:28 am

    Hmmmmm… my hosting is with Bluehost and I haven’t had issues – though I guess I don’t have anywhere near the traffic levels of here.

    The site is very slow for me too so far this morning.

  9. lprent 9
    30 June 2009 at 11:36 am

    Fixed the speed issue (I think). Apache was set to some very sub-optimal levels for accepting clients

  10. jarbury 10
    30 June 2009 at 11:37 am

    Yeah definitely faster now.

    Back to normal, I might even say. (Touch wood)

  11. lprent 11
    30 June 2009 at 11:39 am

    I’ll have a look back in a couple of hours. Have some code to dig through now.

  12. Chris S 12
    30 June 2009 at 11:40 am

    lprent,

    I don’t know much about your setup, but if you want to host overseas I suggest investing in VPS’s.

    You can get a few… one with a bunch of RAM to run your database, one or two to run your webservers and get the host to backwire them with private networking so it doesn’t count towards your cap.

    I don’t know what you’re paying for your colo but 3 vps’s will be < $100usd/mo by a lot and you get full root access to your slices.

    Try http://www.linode.com/ some good specs and good prices. If you do, provision them in the Atlanta data center (I get get a solid 1MB/S+ to my vps although latency is higher).

    Feel free to email me if you want a hand… I do this for a job and would be more than happy to help out 🙂 chris@nevermind.co.nz

  13. infused 13
    30 June 2009 at 11:47 am

    VPS are crap. For one reason, disks. They can be shared with like 10 users. Try running a database on raid 6 or 10 with 10 or so other ‘servers’ on it. Doesn’t work. Been there, done that.

    • Chris S 13.1
      30 June 2009 at 2:48 pm

      With the relatively small amount of data used by a blog, the database can keep most of it cached in RAM.

      • lprent 13.1.1
        30 June 2009 at 3:33 pm

        It does. Most of the 3000 odd posts and 100k comments aren’t accessed except by search-bots and people finding results off them. The bulk goes on 30 posts at the front and the thousand or so comments on those.

        However generating pages is a big exercise on a site that keeps having comments and posts added.

  14. SeaJay 14
    30 June 2009 at 12:01 pm

    Your efforts are much appreciated Standards.
    Who knew so much goes on behind the pages huh?

  15. The Baron 15
    30 June 2009 at 1:30 pm

    Nope, still ultra slow on pretty much every device I visit on.

    • lprent 15.1
      30 June 2009 at 1:59 pm

      I’ll keep tuning it. Only finished shifting it at about 3am. Too many damn graphics

      • The Baron 15.1.1
        30 June 2009 at 3:26 pm

        Yeah, good luck – substantially improved over the last couple of hours.

        Was hanging on the blackberry and on desktop +firefox, now almost back to normal.

        • lprent 15.1.1.1
          30 June 2009 at 3:35 pm

          Bit more tweaking to do. I managed to stop the services when I logged in on the web console. Guess I’d better play with that later. Besides I can do everything ssh

  16. jbc 16
    30 June 2009 at 1:37 pm

    I’ll second Chris S in that VPS are definitely worth investigation if you ever look to re-home again.

    Like everything else there are good ones and bad ones, but a well-managed VPS host will generally offer far better “bang for the buck” than just about anything else (including a bare-rack colocation service). For one thing: you don’t have to worry about managing the metal 10,000 km away.

    The most important thing by far is how clueful the people are who run your service (as you have discovered).

  17. lprent 17
    30 June 2009 at 1:57 pm

    This one is a VPS – they just have a different name for it. The site was also on a VPS from Feb 2008 to March 2009 in NZ with a memory upgrade on the way through from 256MB to 512MB RAM.

    It got moved offshore because of cost (web hosting was about 10x cheaper than the NZ VPS and far more capable than any webhost I could find here). A secondary reason is that makes it harder for someone to use the threat of litigation to shut down the site. It forces the expenditure of bigger bucks up front by the litigant to force the host site to topple.

    A website really doesn’t do much database activity, it is maybe 10% of CPU, and little gets done on the disk. Most of the ‘cost’ is generating the pages from the data and compressing them before transmission.

    Unlike the webapp I wrote for campaigning. Now that has some serious database usage.

    infused: most of the VPS’es I was looking at had separate HDD’s per client.

  18. felix 18
    30 June 2009 at 3:57 pm

    Hey Lynn, just noticed that when I click a “recent comment” link to a comment on the currently loaded page, the whole page reloads instead of just jumping to the comment as it used to.

    • lprent 18.1
      30 June 2009 at 7:30 pm

      I’ll add it to the buglist (Anita keeps updating it). But not tonight…

      • felix 18.1.1
        30 June 2009 at 7:38 pm

        Yes, her attention to detail can be staggering 🙂

        I wasn’t raising it as a user bug as such, just thought it might be a (small) waste of your data and/or cpu cycles.

  19. George Darroch 19
    30 June 2009 at 4:11 pm

    Bluehost have a worsening reputation for doing things like this. They’re also likely to pull your site at the first sign of a cease and desist letter, or any such legalese. See this example regarding Zimbabwe blogs

    I’m glad you’re away from them, and wish you trouble-free hosting!

  20. Madeleine 20
    30 June 2009 at 5:42 pm

    I’m glad you are back – I did wonder what was going on.

    Sounds like you need to keep searching for another host though.

    I hate coding nightmares like that though – I really felt for you reading through it all. Our comments feed died just over a months ago and I have gone around in circles trying to solve the issue staying awake for hours and wasting time trying to get answers out of Blogger. I have now officially given up and we are going to be migrating ourselves to a hosted WordPress platform as a result – thankfully we have been donated free hosting so we shouldn’t have your issues.

    • lprent 20.1
      30 June 2009 at 7:47 pm

      Have fun. WordPress is great – lots of flex.

      The problem in this site is that we get a hell of a lot of comments on a lot of posts. Therefore a lot of page views by both readers coming in from the searches and by the various search bugs.

      The front page isn’t too much of a hassle despite the updates from comment counts. It gets the majority of views, but caches reasonably easily. The speed that comments are added isn’t a real hassle.

      The pages getting commented on was why I put a time limit of about a month on leaving comments on the posts at the end of last year. That immediately reduced the load because those could be reported to the search engines as essentially being static.

      However the search engines are all over any page that has comments being updated. I think that googlebot is the most assidous readers of this site. There are a proliferating number of bot’s. They kind of provide a background noise for the whole site (but we ignore them for reporting) which is why our page views are down compared to some sites. We encourage it by having a sitemap plugin that actively tells a number of search engines about what areas are ‘hot’.

      However I think I’ll have to start discouraging anyone who isn’t a major SE.

      Then of course there are the RSS and Feedcatchers, etc. They are restricted to the last 15 posts..

      Anyway, the point is that the actual commentators aren’t actually the real issue. It all those observing them…. 😈

      Boots the traffic up pretty damn high as a base load. Looks to me like we’re going to have to look at some kind of dedicated systems shortly. That means it will move from the kind of penny-pinching that we’ve run it on to date.

  21. Noddy (used to be Dr.No) 21
    30 June 2009 at 6:32 pm

    Classic. Full marks for introducing Crosby Textor references to an issue that is simply about your inability to manage a website. You guys crack me up sometimes…

    IrishBill: And that’s you on your first warning. Troll again and I’ll ban you for as long as I feel is suitable.

    • lprent 21.1
      30 June 2009 at 7:57 pm

      It was relevant, If there weren’t litigatous fuckwits like Crosby-Textor around the local political sphere, then we’d just dig out some local sponsorship for the servers.

      However the hardware sponsors would be the first target for anyone wanting to use the courts to silence opposition websites. Just as they have tried with Media7 (read the link)

      It isn’t purely a technical issue.

      Anyway, I think you have the wrong name – should be Noddy. Ummm now it is…….. everywhere through the site. Generally it does pay to annoy a tired sysop 😈

Links to post

The server will be getting hardware changes this evening starting at 10pm NZDT.
The site will be off line for some hours.