Autarchy of the Private Cave

Connecting remotely via ssh to my Debian box at work, I needed to mount a CIFS (samba) share, but didn’t remember server name (or IP) and share name.

At least two convenient utilities are available in Debian Lenny for non-X Samba browsing.

smbtree (part of smbclient package) will list all visible workgroups, their servers, and share names of those servers – including “hidden” shares like C$, IPC$, ADMIN$, print$. Very handy and greppable!

samba-commander (smbc package) is a ncurses samba browser with “find file” functionality.

Today, setting up a relatively serious (in CPU resources needed) web-system, I ran into a weird problem of python scripts ending prematurely. After some investigation, it looked like any process which uses up more than 20 seconds of CPU time, is automatically killed. To verify this, I wrote an infinite loop in C,

int main () {
unsigned int i;

for (i = 0; i < 2 ; i++ ) { i = 0; } return 0; } [/c] compiled it and executed several times on the GoDaddy shared hosting server. I did observe the program running for the maximum of 20 seconds of CPU time, not a second more. Please note, that 20 seconds of CPU time can be much more of “real” time, if the script isn’t using 100% of CPU, which often the case for shared hosting. Thus if you have in your php.ini max_execution_time set to, say, 60 seconds, your php script may actually execute as long as one minute; but I’m pretty sure that if your script has lots of CPU-intensive procedures, then as soon as it uses 20 seconds of CPU time, it will be terminated (however, this statement still needs checking – anyone?). To verify, I also created a cron job with the same file. It ran for 30 seconds CPU time. Strangely, this behaviour is not documented anywhere. This limit may also explain a number of other problems, if you have heavy web-applications: they just might be killed before they are finished, causing errors. I do understand the reason for this limitation, and am sure similar limitations exist in other shared hosting environments. The only important thing here is that this limit should have been documented and even put upfront somewhere in the hosting plans descriptions. I also wonder if the limit is the same for all godaddy shared hosting plans, or if it differs. 20 seconds when executed from PHP, and 30 seconds when executed as a cron job were observed on the Deluxe Linux Hosting plan. Extensions, additions and comments are welcome.

Note: have a look at other hosting options.

Update 7: I’m now quite satisfied with page generation times. See other updates at the end of the post and comments to find out more.
Update 8: this blog is not at godaddy anymore for the reasons which have nothing to do with speed.

I’m currently using GoDaddy shared hosting plan. I noticed that my blog, as it grows in popularity and visitors, displays a wide range of page-response times. The best I had seen so far was below 3 seconds of page generation time (note: _not_ page loading, but page generation). If it were the average, I would be happy. However, much more frequently observed times are in range of 20-30 seconds per page. Sometimes pages even timeout, as my uptime tracker service is reporting (For February, there were 100 minutes of unresponsive pages, for March – already over 10 hours!!!).

For example, today around 16:00 GMT the following statistics were reported by my blog:

22 queries.
33.416 seconds.

Evidently, this is unacceptably slow.
Read the rest of this entry »