Skip to contentSkip to conversation

linux and high volume web sites

DISCUSSION valery 4 participants 4 messages Apr 28, 2001 – May 1, 2001

lore
Share & tools
lore

valery

author
#0 lore raw

I have a high volume web site under linux :
kernel is 2.2.17
hardware is 5 bi-PIII 700Mhz / 512Mb, eepro100
all server are diskless (nfs on an netapp filer) except for tmp and swap

dispatch is done by the Resonate product

web server is apache+php (something like 400 processes), database backend is a mysql on the same hardware

in high volume from time to time machines are "freezing" then after a few seconds they "reappear" and response timne is

how can I investigate all these problems ?

#1 lore raw

watch the resonate heartbeat and see if it is getting lost in the network traffic (the resonate logs will show missing heartbeat packets). think seriously of setting the resonate stuff to run at a higher priority so that it doesn't get behind.

depending on how high your network traffic is seriously look at putting in a second nic and switch to move the NFS traffic off the network that has the internet traffic and hearbeat.

I had the same problem with central dispatch a couple years ago when first implementing it. the exact details of the problem that I ran into should have been fixed by now (mostly having to do with large number of virtual IP addresses) but the symptoms were the same.

David Lang

On

Quoted 27 lines · valery

Date: Sat, 28 Apr 2001 11:01:19 +0200
From: valery <[email protected]>
To: linux kernel <[email protected]>
Subject: linux and high volume web sites

I have a high volume web site under linux :
kernel is 2.2.17
hardware is 5 bi-PIII 700Mhz / 512Mb, eepro100
all server are diskless (nfs on an netapp filer) except for tmp and swap

dispatch is done by the Resonate product

web server is apache+php (something like 400 processes), database backend is a mysql on the same hardware

in high volume from time to time machines are "freezing" then after a few seconds they "reappear" and response timne is

how can I investigate all these problems ?
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

#2 lore raw
Quoted 6 lines · David Lang

watch the resonate heartbeat and see if it is getting lost in the network traffic (the resonate logs will show missing heartbeat packets). think seriously of setting the resonate stuff to run at a higher priority so that it doesn't get behind.

Quoted 8 lines

depending on how high your network traffic is seriously look at putting in a second nic and switch to move the NFS traffic off the network that has the internet traffic and hearbeat.

I had the same problem with central dispatch a couple years ago when first implementing it. the exact details of the problem that I ran into should have been fixed by now (mostly having to do with large number of virtual IP addresses) but the symptoms were the same.

In addition to the above make sure there's enough bandwidth to the filer (eg- good switches, multiple ethernets).

Consider moving to 2.2.19. Significant VM changes after 2.2.19pre3 which could account for the freezes.

rgds,
tim.

Quoted 7 lines
Quoted 7 lines

I have a high volume web site under linux :
kernel is 2.2.17
hardware is 5 bi-PIII 700Mhz / 512Mb, eepro100
all server are diskless (nfs on an netapp filer) except for tmp and swap

dispatch is done by the Resonate product

Quoted 8 lines
Quoted 8 lines

web server is apache+php (something like 400 processes), database backend is a mysql on the same hardware

in high volume from time to time machines are "freezing" then after a few seconds they "reappear" and response timne is

how can I investigate all these problems ?

--

#3 lore raw
In reply to David Lang

Also make sure you aren't suffering database lock contention from Mysql. This causes very fast context switching on the database server, and is typically unable to do useful work even though its load avg is not high. "vmstat" is useful here.

Sean

Quoted 48 lines · Tim Moore
Quoted 6 lines · David Lang

watch the resonate heartbeat and see if it is getting lost in the network traffic (the resonate logs will show missing heartbeat packets). think seriously of setting the resonate stuff to run at a higher priority so that it doesn't get behind.

Quoted 8 lines

depending on how high your network traffic is seriously look at putting in a second nic and switch to move the NFS traffic off the network that has the internet traffic and hearbeat.

I had the same problem with central dispatch a couple years ago when first implementing it. the exact details of the problem that I ran into should have been fixed by now (mostly having to do with large number of virtual IP addresses) but the symptoms were the same.

In addition to the above make sure there's enough bandwidth to the filer (eg- good switches, multiple ethernets).

Consider moving to 2.2.19. Significant VM changes after 2.2.19pre3 which could account for the freezes.

rgds,
tim.

Quoted 7 lines
Quoted 7 lines

I have a high volume web site under linux :
kernel is 2.2.17
hardware is 5 bi-PIII 700Mhz / 512Mb, eepro100
all server are diskless (nfs on an netapp filer) except for tmp and swap

dispatch is done by the Resonate product

Quoted 8 lines
Quoted 8 lines

web server is apache+php (something like 400 processes), database backend is a mysql on the same hardware

in high volume from time to time machines are "freezing" then after a few seconds they "reappear" and response timne is

how can I investigate all these problems ?

--
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/