Tip-Top Dreamhost, a ‘load’ of overselling?

One of my boxes is actually a Dreamhost account. I use this box with the promise of high bandwith and large storage for low annual cost. I really do take advantage of this offered service. However, I went to log in and was greeted with a rather sluggish shell. I ran top and got “427.38” (!) as current load (for those that are not aware – it ideally shouldn’t really exceed 1.00). It’s probably the highest load I have ever seen on any *nix box – but still provide a usable (but very slow) SSH connection and yet recover back to a reasonable load.

To be fair to Dreamhost, it did recover itself within a few minutes – but I was quite surprised at this load and thought it worthy of mention.

Screenshot of “top”:

Dreamhost “top” - high load

October 24th, 2007 | All, Blogroll, hantslug, ubuntu, ubuntu-uk

7 comments

This is exactly the reason why I left DreamHost not once, not twice, but three times. They wildly oversell, the owner brags about out in the open see their blog. Things frequently just don’t work. I cannot tell you how many times mail service was down or my sites were down completely.

For $10/mo what do you really expect? It says something about a company when they continue to over sell when they cannot even support their current customers.

I gave ‘em 3 separate chances on 3 separate servers, all the same bull DB powered sites (most everything I work with) bone ass slow. So, I upgraded my current VPS and moved all my sites there, I’ve been sailing along happily since. :)

Comment by Big Dan — 28 October, 2007 @ 1:26 am

This hit me a lot as well. Chances are, at some level dreamhoststatus.com or a CS-rep will tell you it’s a Filer issue. Every time my uptime numbers came back in the double or triple digits, they blamed their network Filers.

I chose (and very happily so) to move to MediaTemple for my hosting – I also went to their DV over their GridService, so it’s not apples to apples, but Dreamhost’s sardonic too-cool-for-school attitude was just wearing me down – and it was always more of the same – latency on the filers, and not nearly enough redundancy for even my piddly little personal site. I’m not asking for corporate level uptimes, but 90% would be nice (that’s not asking much)!

Comment by R.J. Lorimer — 24 October, 2007 @ 3:39 pm

Actually, in some unix environment (ie. Solaris) load IS relevant to number of processes. 100% CPU usage in 2 CPUs would be 2.00. :)
R.

Comment by Damian Wojslaw — 24 October, 2007 @ 3:21 pm

<p>Top show 3 processess… maybe some accounting fluctuations made load so big?</p>

<strong>Author Comment:</strong>
These are the three processes that user ‘daviey’ is running. I have more than one account on this box – and there are other customers.

Comment by zdz — 24 October, 2007 @ 1:58 pm

Recipe for having system load in the thousands: have a cron script that records some system statistics with rrdtool every five minutes, have NFS mounts, disconnect the Ethernet cable and leave it for two days.

Many thousand cron scripts that were all blocked waiting on the dead NFS mount suddenly come alive at once. They try to send many thousand error messages vial mail to root, all at once. When you finally manage to ssh in, you see 15-minute loadavg over 4000.

Fun!

Comment by Marius Gedminas — 24 October, 2007 @ 1:48 pm

Isn’t the load average relevant to the number of the processor/core in the system? Maybe the system have 430 processors :)

Author Comment:
I like your logic!
$ cat /proc/cpuinfo | grep name
model name : Dual Core AMD Opteron(tm) Processor 175
model name : Dual Core AMD Opteron(tm) Processor 175

Comment by Ercin EKER — 24 October, 2007 @ 1:24 pm

I once had loads that high on Dreamhost. I shot them an email and they offered to move me to a hand picked server with low load while they resolved the issue. Since then my sites have been humming along.

Comment by Benjamin Golub — 24 October, 2007 @ 1:09 pm