News Feed
Recent Blog Posts
Recent comments
That's great, thank you.
Sorry I didn't go through and update all my old...
Is RbMan2 still available anywhere? The link...
Slicehost is very stable and professional and the...
I enjoyed reading your article, and the followup...
That's great, thank you.
Sorry I didn't go through and update all my old blog posts when I moved it. RbMan is available on riaforge at the following URL: http://rbman.riaforge.com/
Is RbMan2 still available anywhere? The link doesn't work now.
Slicehost is very stable and professional and the support is great. I'm still running my jumba server for some thing, but their support is not great. The problem I was having was caused by some networking problems with some package updates. The jumba support guy tried to tell me it way my problem because I had run aptitude update .... seems like an incompatibility with parallels to me.
Anyhow, I re-imaged and am running it without updating.. Seems to be putting along ok atm.
I enjoyed reading your article, and the followup comments.
I've been a slicehost user for quite a while now, and have been supporting a small list of wordpress clients on it as well, and I've been mostly satisfied with the reliability and performance over the last 2 years, but I've also been a bit curious to see what else is out there.
From your notes, it seems I've made a decent choice all around...
Nothing magical, just the Android web browser ... I'm using Drupal for my blog.
How did you do it? Was it using the Android browser on your site, or is there an App?
I was thinking about putting together a mini mobile blog app that could connect to Blog CFC, so am intrigued.
Hi,
We use AbleCommerce for CFMX and have gradual memory exhaustion. Is there a known fix for AC on CFMX?
Thanks,
Tony
It's now been 48 hours and another email from me to the support people at Jumba and still no reply ... my confidence is waning.
I've just rebooted my jumba server again, this time through the parallels management console and the server is now unresponsive....again. The parallels console says it's running,but it won't respond to ping or ssh. I hope their "technical support" comes back with a response about this issue soon.
I just signed up for a server at jumba.com.au. As mentioned previously they have some good looking offers. I ran the same set of performance tests on the server and came up with comparable performance to slicehost. The sign up process was quite smooth and the management console is quite feature rich. It's still not quite up the the standard of slicehost, but has everything you really need.
I did run into one minor hickup so far. After rebooting the server on the command line it became un-responsive. A help desk request on the issue has been more than 24 hours without any response, so I ended up re-installing the OS through the management console (Parallels). This is far from desirable. I'll see how it goes with a little more prodding and poking ...
=== jumba.com.au 768mb ===
FS-Mark (Disk I/O): Average: 15.4 Files/s
AIO-Stress (Disk I/O): Average: 87.07 MB/s
Apache Benchmark: Average: 2659.57 Requests Per Second
Timed Apache Compilation: Average: 178.47 Seconds
RAMspeed: Average: 1900.94 MB/s
I've continued to experience memory and stability issues with the burst.net server, so I've decided to drop them as an option... I knew it was too good to be true.
Using wildcards in the host name might be causing a problem. Here is an example of a working config.
{5526, ejabberd_service, [{ip, {127, 0, 0, 1}}, {access, all},
{hosts, ["bandersnatch.home"],
[{password, "secret"}]}]},
You need this in your modules section and of course the bandersnatch daemon needs to be running.
{mod_service_log,[{loggers, ["bandersnatch.home"]}]},
Sure thing. Running ejabberd 2.0 with mnesia. Seems to be that ejabberd isn't starting when I add the lines:
{5526, ejabberd_service, [
{ip, {127, 0, 0, 1}},
{access, all},
{hosts, ["bandersnatch.**myhost.com**"],
[{password, "secret"}]
}
]},
I can't find anything in the ejabberd.log to show reason for not starting but I get no connection and the web admin doesn't come on. Is there any other log to look at other than /var/ejabberd/ejabberd.log?
Hi JD, If you could post some more information it might shed a bit of light on the issue. There's not much to go on there.
[11:52:05] Bandersnatch: (ERROR) Unable to connect to Jabber server (localhost) ...
[11:52:05] Bandersnatch: (HASH(0x166d750))
Any ideas???
As part of my testing I thought I play around with the burst.net server a bit and install cold fusion. Surprisingly, as in I've never seem this kind of error before. (I've been using coldfusion since 1996) The system imploded while I was running a $tail -f on the coldfusion log and then seemed to magically come good when I stopped the tail.
After running a few simple commands while the tail was running on another console following errors were returned.
root@vps1:~# ps -A
-bash: fork: Cannot allocate memory
root@vps1:~# ls
-bash: fork: Cannot allocate memory
root@vps1:~# htop
-bash: fork: Cannot allocate memory
root@vps1:/opt/coldfusion8/bin# ls
Segmentation fault
root@vps1:/opt/coldfusion8/bin# cd ..
Segmentation fault
Also, interestingly the coldfusion server I was testing output the following in the logs.
Java HotSpot(TM) Server VM warning: Attempt to deallocate stack guard pages failed.
I'm going to look into this and will report back, but I just couldn't be bothered right now. It's Friday night after all ....
The control panel they have is great from the perspective of overall OS/Disk tasks (they don't have any general installed services management panel e.g like plesk/cpanel - you'd have to install that yourself if that's what you need).
Should be pointed out that they're an unmanaged service, so they won't provide any hand holding beyond core functionality. I don't recall ever having to get in touch with them, one time I knocked my box offline (my own fault) I got help in their irc chat in 5 minutes.
I really like the look of Linode, especially their stack scripts. I think I'll give them a test in the next few days. The other one I have my eye on is an Australian provider called Jumba. They've just released what look like some good packages. More details here:
http://www.linode.com/features.cfm
http://blog.linode.com/2010/02/09/introducing-stackscripts/
http://www.jumba.com.au/vps
It took them a couple of days, but burst.net have re-sized my vm to 1Gb of ram and 1.5Mhz CPU. It seems it is a manually process they're doing when creating and resizing. Also, it was only through prompting them through their support desk that it got done.
Anyhow, I've run the tests again on the new VM and the results show further improvements in some areas, but it's a little inconsistent with what I would expect. i.e. With more CPU available you'd think apache requests would increase. Disk I/O bandwidth is up and file throughput is down. I don't really get that, but it may have something to do with the reduced ram speed. I probably could do some more thorough tests, but this is all I have time for atm.
=== burst.net 1024 ===
FS-Mark (Disk I/O): Average: 73.40 Files/s
AIO-Stress (Disk I/O): Average: 517.44 MB/s
Apache Benchmark: Average: 4793.29 Requests Per Second
Timed Apache Compilation: Average: 118.72 Seconds
RAMspeed: Average: 2613.66 MB/s
Thanks for the insight Mike. I can't explain the huge performance difference between the two. It may be that I've started out on an empty box ... just guessing. The CPU looks better on the burst server, but I can't imagine they'd be providing raid 10 for those prices. It's a strange paradox. One question ... how do you find the support and management tools a linode?
You also need to take into account the core virtualisation platform, and the likely quality (and hence cost) of infrastructure. Slicehost ar a xen based provider where burst.net are openvz based.
The reason burst.net are so cheap is that they are using openvz as the basis of an 'oversold' service (essentially betting against all users simultaneously requiring maximum resources). Xen vps have greater isolation between users, and cannot be oversold to the same extent, it has a higher overhead but provides a more consistent level of performance.
My experience with both openvz/xen based systems has shown that xen systems have far less periodic performance issues and far greater uptime. If you want a bit more for the money look at linode who are a similar quality host to slicehost.
Burst.nets incredibly cheap prices also undoubtably attract a lot of inexperienced, unsuitable and dubious users (have a browse through webhostingtalk to see some of the burst related tantrums to see what I mean), which ultimately will contribute to greater problems through network loading, DOS, CPU hogging etc. I wouldn't personally entrust anything critical to that kind of network neighborhood, although as a part of a distributed system it's undoubtably a cheap source of bandwidth.
I found this discussion on lowendbox.com that seems to support my conclusions so far.
http://www.lowendbox.com/blog/burstnet-5-95-openvz-vps-with-512mb/
that's really a fantastic post ! added to my favourite blogs list.. I have been reading your blog last couple of weeks and enjoy every bit. Thanks.
Hi Dan,
I couldn't find it on the bandersnatch home page, but found it through a mirror site or something. Can't remember the link off hand. I've attached a copy here so you could use that one if you want.