Occasional downtime... - Page 2

richandzhaoyan · 20th February 2007, 09:11 PM

Tim,

Do you know if the downtime you see is regular?

I did have a similar problem a couple of years ago where my site, including ftp and email went down at midday and midnight for about 2 minutes every day. The cause of this was found to be the log rotation (for the server, not just my site).

I dont know if the problem could be related as my site is now on a different server (upgraded package) and when the problem was occuring the site was on one of the US servers. I would imagine the new UK servers would not take as long for the log rotation.

Rich

CSNM-Carl · 20th February 2007, 09:33 PM

Quote:

Originally Posted by richandzhaoyan

Tim,

Do you know if the downtime you see is regular?

I did have a similar problem a couple of years ago where my site, including ftp and email went down at midday and midnight for about 2 minutes every day. The cause of this was found to be the log rotation (for the server, not just my site).

I dont know if the problem could be related as my site is now on a different server (upgraded package) and when the problem was occuring the site was on one of the US servers. I would imagine the new UK servers would not take as long for the log rotation.

Rich

Oh this is interesting? Although when I can't get to any sites on my server I can't access any sites on a friends server with UH either. Do all the logs on the servers rotate at the same time? UH?

percepts · 20th February 2007, 09:35 PM

I beleive that log rotation takes place in the early hours of the morning when most people are asleep. You can see when it happens by looking at your raw log files to see at what times the first and last entries in each daily file occur. That tells you the time the file was closed/opened assuming there were visits to your site at that time. Checking a few days files will give you the approx time it happens.

thinking about it a little further, you can ftp to your log file folder and see the times the log files were generated which is an easier way to do it.

MrBen · 20th February 2007, 09:41 PM

Log rotation is an option in the site admin control panel. It can be either 4 times Daily; Daily; Weekly. The default is daily and for me this is around 5-6AM GMT.

Ben

UH-Tony · 20th February 2007, 09:43 PM

The log file rotation occurs around 4:20-4:40am across all servers (specific to the servers timezone).

It use to be that log rotation would run 4 times per day if even a single site had set their log rotation to occur 4 times per day vie the option in the control panel however we disabled that feature for all shared/reseller servers due to the fact that it would cause Apache to be reloaded 4 times daily (it is required to complete the log rotation process).

percepts · 20th February 2007, 09:43 PM

when you do your traceroute are you specifying a timeout value ?
You should not need to specify a high timeout value for the tracert. If you need to specify one then that should be telling you that there is a problem. And if you are specifying a timeout then please run the traceroute again without a timeout period and post the results.

Simon · 20th February 2007, 10:15 PM

I'm in Spain and I get results such as this

Code:

# traceroute ventura.unitedhosting.co.uk
traceroute to ventura.unitedhosting.co.uk (83.223.124.13), 30 hops max, 38 byte packets
 1  192.168.0.1 (192.168.0.1)  0.223 ms  0.212 ms  0.161 ms
 2  cliente-26627.iberbanda.es (83.230.168.1)  50.319 ms  49.467 ms  50.813 ms
 3  cliente-17863.iberbanda.es (83.230.133.197)  89.490 ms  89.126 ms  59.883 ms
 4  madr-m202.red.iberbanda.es (217.11.96.93)  54.930 ms  70.295 ms  84.221 ms
 5  87.111.24.218 (87.111.24.218)  60.231 ms  49.933 ms  44.233 ms
 6  80.91.67.245 (80.91.67.245)  55.236 ms  50.936 ms  48.963 ms
 7  * * v3492.mpd01.mad04.atlas.cogentco.com (130.117.2.58)  52.601 ms
 8  interoute.espanix.net (193.149.1.42)  54.442 ms  43.627 ms  64.986 ms
 9  PO6-0.mad-001-access-1.interoute.net (89.202.161.1)  75.241 ms  89.829 ms  89.720 ms
     MPLS Label=225 CoS=5 TTL=1 S=0
10  PO10-0.par-gar-core-1.interoute.net (84.233.204.233)  100.450 ms  94.381 ms  94.237 ms
     MPLS Label=282 CoS=5 TTL=1 S=0
11  PO1-0.lon-wal-core-2.interoute.net (84.233.152.165)  110.014 ms  114.465 ms  85.275 ms
     MPLS Label=316 CoS=5 TTL=1 S=0
12  Te9-1.lon-wal-access-4.interoute.net (217.118.119.26)  135.239 ms  108.679 ms  149.815 ms
13  84.233.153.30 (84.233.153.30)  96.066 ms  235.279 ms  158.872 ms
14  ge-0-3-0.core-1.lhc.lon.uk.as29017.net (89.145.125.17)  215.329 ms  284.345 ms *
15  ventura.unitedhosting.co.uk (83.223.124.13)  80.645 ms  105.486 ms  84.062 ms

As mentioned before, try and connect your PC directly to the internet and skip the router and then report back.
Local problems can be down to all sorts, a cable that got trapped behind a desk and crushed is probably the most commonm, or just old equipment.

Is you website high traffic? If it is then maybe the log files could be causing the problem, but then you'd see the problem at the same time everyday. You seem to be sayign that its more a random event.

As Tony said its a good idea to fix what you can fix first and then work with whats left, those early timings are very high

TygerTyger · 20th February 2007, 10:19 PM

Yes, I've noticed this occuring on my server (delta). Namely very short blips of inaccessibility, perhaps 15 seconds to a couple of minutes at the most. The only site I'd notice it on is a forum, so plenty of other people to corroborate the happenings, which they have done, and I have other sites on the server so I can check that they aren't loading either during the downtime, which they are not.

I've never paid much attention though, the service is flawless the rest of the time and the downtime is extremely short. But uh, in light of this maybe I'll make a note of if/when it happens again. I just figured it was someone running a heavy script somewhere. It may well be regular, I hadn't thought about it.

timcclayton · 20th February 2007, 10:44 PM

I will head off and check out my settings, swap in another router/cable etc. to see if I can reduce the initial connection delay. I am confident that this is absolutely not the issue, but at least it might make UH sit up and take the problem a bit more seriously. What I want is not heads on a platter, but a resolution so that we can eliminate this issue.

It is interesting that quite a few people seem to be having this issue though, so thanks for posting (and anyone else who reads this please feel free of course to add comments - it all helps).

I will report back once I reduce that initial delay.

vord · 20th February 2007, 10:56 PM

I have exactly the same problem. The server runs fine but I can't get there all the time. When the UH forum was hosted on US servers I'd come on here to check server status, but I guess the forum is UK hosted now?

It's not just my sites. I've noticed other sites elsewhere on the internet don't work at the same time, though Google and a US site I frequent do work while I can't make it here. Network issue on the way somewhere on the way is my guess, but it would be interesting to know what it is.

My working theory - some router on the way busy from time to time causing problems with traffic towards London?

UH-Tony · 20th February 2007, 11:05 PM

Actually these forums are still hosted in Houston, TX USA.

percepts · 20th February 2007, 11:06 PM

I have also had delays from time to time but it is difficult to determine the cause.

The point about doing a tracert is that it may show timeouts along the route to the UH server. BUT the route taken is not guranteed to be the same each time so even tracert is not guaranteed to show where the problem lies. If your request for site is being timed out along the route then you won't get your page and the slower your connection the the more likely you are to be timed out along the way. At least thats how I understand it. Since your connection seems to be at least 7 to 8 times slower than a 56k dial up connection, I would suggest that you are at a much higher risk of being timed out than the vast majority of people.

The fact that google or any other site works means nothing since the route is very probably different than the route to UH.

Having said all that, if the timeout happens only one or two hops before UH then its unsurprising that everywhere else seems to be working OK. i.e. you cannot say the fault is the UH server with any confidence of accuracy.

So, fix your connection speed problem and see if that fixes the access problems.

Simon · 20th February 2007, 11:15 PM

Google is a very poor example in this case. They have servers all over the world and i'm sure we are alwys routed to the best available at that particular time.
Try and comare speed/connections with something that is likely to be hosted in a single location.

Simon

UH-Matt · 7th March 2007, 09:55 AM

I have posted a summary of the issue we believe exists for NTL users and will keep you updated here:

http://unitedforums.co.uk/vb/showthr...0382#post60382

Feel free to contact the helpdesk if your on NTL and have any concerns.,