Since a few days I'm experiencing a sudden drop of http-traffic every hour or so, for a few minutes and then everything comes back up, without any error in any of the logs and whatsoever. After a lot of looking, the only logical conclusion from my part was that it HAD to be external (sorry if it aint and I'm missing it, but still...). The server normally has 100+ request per second, quasi all day long, so these drops (see attached image, above the red dots) are very unusual.
So I decided to create a trouble ticket:
QUOTE
10/14/2004 3:40:41 AM
Since a few days, it seems that every X number of minutes our server suddenly no longer receives webrequests during 2 or 3 minutes.
We are not the only ones since someone we know that has multiple EV1 accounts too is experiencing the same problem. Is there anything wrong in the datacenter or something?
Since a few days, it seems that every X number of minutes our server suddenly no longer receives webrequests during 2 or 3 minutes.
We are not the only ones since someone we know that has multiple EV1 accounts too is experiencing the same problem. Is there anything wrong in the datacenter or something?
Support answer:
QUOTE
10/14/04 4:10:25 AM
WebTech
Dear Customer,
We didnt notice any problem while investigating you server, we have checked the logs, messages, dmesg. While investigating /var/log/httpd-error.log we got mod_gzip: TRANSMIT_ERROR:32 which means :
Whenever you see TRANSMIT_ERROR:xx. the 'xx' is the actual Operating System error that happened while mod_gzip was 'sending' something back to the client.
Error 32 is 'EPIPE' which means 'Broken Pipe'. That's the error that is always received by a Server when the client 'goes away' while the Server is sending data.
It usually just means the user pressed the STOP or BACK button on their browser and abandoned the download. It's a normal error. Current version of mod_gzip logs ALL error codes for now so error 32 can be safely ignored. It just means the browser 'bailed out' during a download
Rest we didnt fine any problem in it. Please confirm once again, if you still face the problem do contact us.
Thank You,
Ev1servers Support.
WebTech
Dear Customer,
We didnt notice any problem while investigating you server, we have checked the logs, messages, dmesg. While investigating /var/log/httpd-error.log we got mod_gzip: TRANSMIT_ERROR:32 which means :
Whenever you see TRANSMIT_ERROR:xx. the 'xx' is the actual Operating System error that happened while mod_gzip was 'sending' something back to the client.
Error 32 is 'EPIPE' which means 'Broken Pipe'. That's the error that is always received by a Server when the client 'goes away' while the Server is sending data.
It usually just means the user pressed the STOP or BACK button on their browser and abandoned the download. It's a normal error. Current version of mod_gzip logs ALL error codes for now so error 32 can be safely ignored. It just means the browser 'bailed out' during a download
Rest we didnt fine any problem in it. Please confirm once again, if you still face the problem do contact us.
Thank You,
Ev1servers Support.
Surely you must be joking? Those 32 errors are pretty darn normal and happen once in a while...
I replied:
QUOTE
10/14/2004 4:48:17 AM
Dear sir,
I know that the problem isn't local, I already checked this out myself (and I know the gzip error, that happens if someone disconnects or closes the browser and is normal). BTW: The actual httpd logs are in /var/log/vhosts. /var/logs/httpd-error.log is simply a collect for all except our other sites and is used rarely.
As i said, we know another person that has 3 EV1 servers that can't even locate each other anymore at some times, he is experiencing EXACTLY the same symptoms as we do.
Are you sure there is no switch problem or whatsoever?
The problem is kind of annoying for the both of us since we both have pretty big sites.
If you look in our logs, generated by your servers at: you can "clearly" see the dropouts, we normally have constant traffic of more than 100 requests per second.
My colleagues logs:
If you look at his stats, you'll notice exactly the same thing there.
All "holes" in there are periods that his servers can't communicate with each other (all EV1 servers and all 3 of them at the same time). Coincedence?
PS: Neither of us has made any configuration changes in that period.
Dear sir,
I know that the problem isn't local, I already checked this out myself (and I know the gzip error, that happens if someone disconnects or closes the browser and is normal). BTW: The actual httpd logs are in /var/log/vhosts. /var/logs/httpd-error.log is simply a collect for all except our other sites and is used rarely.
As i said, we know another person that has 3 EV1 servers that can't even locate each other anymore at some times, he is experiencing EXACTLY the same symptoms as we do.
Are you sure there is no switch problem or whatsoever?
The problem is kind of annoying for the both of us since we both have pretty big sites.
If you look in our logs, generated by your servers at:
My colleagues logs:
If you look at his stats, you'll notice exactly the same thing there.
All "holes" in there are periods that his servers can't communicate with each other (all EV1 servers and all 3 of them at the same time). Coincedence?
PS: Neither of us has made any configuration changes in that period.
The answer (notice the time it took to reply):
QUOTE
10/14/04 11:32:52 AM
WebTech
Dear Customer,
I have inspected and confirmed that wht you are seeing on the mrtg graphs is normal. I have also checked other graphs and the tmie frame that you see the dip in the trafice is normaly not just on your servers but others as well as this is around the time most people go to sleep. I have also checked with our NOC and they have no reported outages for a while at either Data center for bandwidth.
Thank you
Ev1 Servers Support.
WebTech
Dear Customer,
I have inspected and confirmed that wht you are seeing on the mrtg graphs is normal. I have also checked other graphs and the tmie frame that you see the dip in the trafice is normaly not just on your servers but others as well as this is around the time most people go to sleep. I have also checked with our NOC and they have no reported outages for a while at either Data center for bandwidth.
Thank you
Ev1 Servers Support.
I had to read this twice to believe it, but there it was, a statement that my stats dropped because people went to sleep!
I still can't believe it.
PS: For correctness, I have to report that the probleem seems to be solved right now, or at least hasn't occured for the last few hours and that we have been EV1 customers for multiple years now and never had much problems (and if we had one it was solved correctly and fast).
Again, sorry but I had to write this, don't wanna step on anybody's toes, but I really don't like to be threated as being a moron (npi).
Blizz.