(This post is a backup of http://sogno.net/blog/?p=96 .)
After a terrible experience with the hosting of Aruba that I was losing post, he was down too and often had big problems with PHP (all seasoned with care to say the least painful), I decided to change everything and move to a virtual server.
A virtual server is essentially a "slot" built on a real server, the user can do (almost) everything you want.
A friend advised me to contact the service AreaServer of Energit , telephone operator and provider of Sardinia.
I contacted them about the September 10, 2004 and, after a fruitful exchange of information via e-mail, Energit offered me a trial week of service.
I accepted, and I must admit that during that week, everything worked decently, but there is also to say that I had no domain associated with the server so I could not play much. However, their availability seemed positive and I decided to turn my virtual server and to transfer my two main domains (sogno.net and nicolucci.net) with them, in effect giving up the remaining months on Aruba.
The activation was rapid enough, a couple of days, and I immediately noticed something wrong: using the web control panel to add the site www.nicolucci.net, the change was not "noticed" by the system . I immediately called the hotline and the answer was "the rule has not yet been transferred and the system does not see it." The guy was not very safe, and I I was even less know how these things work and it seems foolish to make such a check, because I could easily have a domain managed by an external nameserver that points to the virtual server ... I eventually solved by manually editing httpd.conf.
now is the fourth month of service (October-November-December-January), I pay € 9.90 charged to your monthly bank account and have an annual subscription. Except that the money does not grow on trees and can not afford to throw money to the wind in the following months, I would like to Aruba and I found a better place to manage my stuff. The problems, in fact, are rife.
users do not have available a true root access, and virtualization has made "good."
have a user admin, but root is only pretend, the filesystem is shared and there are only red lines to prevent the spill, and I firmly believe that somewhere there is a leak and also gives access files of other users, although ps aux does not show the processes of others, you can work around this "protection" by doing something like
cd / proc; for i in *; do echo $ i>> ~ / processes; cat $ i / cmdline>> ~ / processes; echo>> ~ / processes; echo>> ~ / processes; done
and then going to read the file processes in their own home. The demons
used on the server are old and buggy and obviously personally can not be upgraded because they do not have the root.
Suffice it to say that there is a version of ipopd3d two years old (v2003.83), which obviously does not like much memory allocation enforced by xinetd. Result: If you receive a message larger than the memory allocated by xinetd, when trying to download mail ipop3d crashes. I also provided assistance for a problem on my analysis (in practice I think that version of ipop3d try to load the entire mail message, and obviously not having enough available memory you get a SIGSEGV), and after a couple of months I was assured that the limit was raised from 8 MB to 32 MB. Thanks a lot, but in the meantime, every time I access via webmail, charged forward messages to a decent box (eg gmail), delete them from webmail, webmail and delete them from the basket and then finally re-download the other messages with my client. Something quick and easy, really.
In response to one of my first contact care about this ridiculous problem, I was told to use ftp to transfer large files, so basically I had to set up a ftp access to receive large attachments. I it is obvious that a solution is impractical: do not always know in advance that someone has to send an attachment of 10 MB.
Among other things, the ftp server has a problem to say the least absurd during upload, the files are truncated to 50 MB accurate, regardless of the space allocated for the user. I've lost an afternoon thinking it was my problem, and I finally resolved by the. Zip file smaller than 50 MB and loading them separately. They were my recordings to mp3, fortunately, so I could do it. But if it was my video production?
Returning to the question, was forced to use sendmail , which is notoriously complex, and majordomo, which is not much better.
I wanted to install Mailman to manage my mailing list, but the problem is the same: not having the root, you can not. Point.
What I'm wondering, and I have also openly said to them, it's like they have a turnover even if a customer who has quietly played with unix us such problems. How does a typical "average user" to use these services? It will be for the (forgive the bad thoughts) that are annual contracts, so, whatever happens, the income is there?
But back to the service description: web statistics are almost unusable, as they relate default, the current month. This means that if the first of the month I want to see what happened the previous month, I have to do something like
/ usr / local / bin / awstats_buildstaticpages.pl-update-config = www.sogno.net-dir = / home / web / www.sogno.net / website / awstats-awstatsprog = / home / httpd / docs / awstats / wwwroot / cgi-bin / awstats.pl-staticlinks-month = year = 2004-12
and go see my page . Subsequent to regeneration, of course, pages of statistics will be overwritten with the latest data.
Again, I think the classic average user does not have the knowledge to carve out a command typically read pieces of script in bash, but may need statistics complete for commercial reasons.
addition often the server is unreachable : sometimes it happens to be active but with all the demons are not working, and all connections are refused because the doors are closed. The solution here is to restart the virtual server at least two or three times through the control panel. Only once is not enough, and nobody knows why.
Other times the machine load is scary . According to / proc / cpuinfo, the machine that is my virtual server is a 2.8 GHz Pentium IV running at 1.4 GHz:
...
model name: Intel (R) Pentium (R) 4 CPU 2.80GHz cpu MHz
: 1399,999
...
and physical RAM is 900 MB with over 5 GB of swap. A car of all respect (except that the CPU runs at half the clock), but more often you get a load of more than 150 days ago and I saw a peak of 260. The "loaded" in question are those shown by uptime.
In these cases the network is working perfectly, because it is managed by the kernel, but everything runs in user-space, ie practically the rest of the software, it is monstrously slow.
This belief makes the perfect idea:
[admin @ vs3089 root] $ time
uptime 21:13:54 up 3:13, 0 users, load average: 159.90, 119.98, 82.68
real 0m41.713s user
0m0.000s sys 0m0.010s
Anyone who has used a bit 'a unix knows that 41 seconds to return the output of uptime is really too much. Just to compare, my old PC (a 850 MHz Duron with 512 MB of RAM), which also has some 'demons and software running, it takes three milliseconds.
The problem on that server is that there are too many virtual servers with too many processes running.
I tried to make a simple approximate calculation. My virtual server was running about 30 processes, which are the "standard" because I have no other process "personal" focus. In / proc there are over 1400 pseudofile. Assuming that they are also only 1300 (to put into account the processes the server's "true") and that each virtual server have exactly 30 processes running (but I know for a fact that is not the case), it turns out on that machine there are more than 43 virtual servers .
And frankly I do not think a machine capable of withstanding a load like that, because, as there is no actual division of the CPU and RAM between different users, you can run any process .
The result is what I have described: loads so high that the connections are accepted but must inevitably time out.
Often, for this problem, I find myself unable to check e for hours . Yesterday, during a blackout like this, I calculated the time required to get an answer from the mail server that the user foo with password foo does not exist and then disconnect. The result:
innocence: jollino ~ $ time echo-e "USER foo \\ r \\ nDo you want foo \\ r \\ nQUIT \\ r \\ n"
assistance then, I'm sorry to say, but is poor and rough. On the phone they are often given answers
dubious, such as that on the control panel that was not updated with new sites or that the 50 MB limit for uploads ftp, and mail is often the answers come only after days, and are equally unconvincing. Even to receive a response to my many email complaints I had to print my message and send it by fax, as
had been a week and I had not received anything
.
In general, then, they also seem a victim of the syndrome of "we're working on," as well as Aruba, and do not need a marketing genius to realize that the promises are maintained not only to feel cheated customers.
Here are my final grades.
Price: 8 / 10 Start: 6 / 10 Features: 4 / 10 Support: 5 / 10 Stability: 3 / 10
Ease of Use: 3 / 10
Speed: 5 / 10
MEDIA : 4.8/10
Judgement: to be avoided. Better to spend a little 'more and have a lot more.
I personally am planning to turn to, when the one-year contract expires at
RimuHosting
, a provider headquartered in New Zealand, but with servers in the United States. I contacted them for information on service and they look fantastic, they're geeks who know what they're talking about (and not the business that are reinvented system administrators) and respond quickly. And these are real virtual server based on User-mode Linux
, which has a real and a real root filesystem private, as well as real allocation of processor and memory.
0 comments:
Post a Comment