Memory problem on Linux 2.6.5

[Clint Byrum]

mod_backhand is showing erroneous results for the amount of memory on a
machine when running Linux 2.6.5.

Note:
Ent Hostname Age Address Total Mem Avail Mem
--- --------- --- ------------- ------------ -----
7 node2 1 1.2.3.5:80 1008 MB 850 MB
11 node3 1 1.2.3.4:80 4294967295 MB 4294967295


node 2 and node3 are identical hardware.. but node3 is running 2.6.5,
and node2 is running 2.4.25. I'm sure this is due to the change in
/proc/meminfo.

On a side note, Arriba also comes out *way* lower on this box running
2.6.5. Maybe the box is just slower running 2.6.5 ... or maybe its the
HyperThreading again. :-P

[Theo Schlossnagle]

Post by Clint Byrum
mod_backhand is showing erroneous results for the amount of memory on
a machine when running Linux 2.6.5.
Ent Hostname Age Address Total Mem Avail Mem
--- --------- --- ------------- ------------ -----
7 node2 1 1.2.3.5:80 1008 MB 850 MB
11 node3 1 1.2.3.4:80 4294967295 MB 4294967295
node 2 and node3 are identical hardware.. but node3 is running 2.6.5,
and node2 is running 2.4.25. I'm sure this is due to the change in
/proc/meminfo.

Rainer has sent me to patches that should have addressed this problem.
I don't have access to a 2.6.x machine on which I can test mod_backhand
at the moment. Are you running the latest mod_backhand from CVS?

Post by Clint Byrum
On a side note, Arriba also comes out *way* lower on this box running
2.6.5. Maybe the box is just slower running 2.6.5 ... or maybe its the
HyperThreading again. :-P

Perhaps... you can manually set the arriba by editing the file it
writes. The arriba is a very arbitrary metric, so candidacy functions
and people should take it with a grain of salt.

// Theo Schlossnagle
// Principal Engineer -- http://www.omniti.com/~jesus/
// Postal Engine -- http://www.postalengine.com/
// Ecelerity: fastest MTA on Earth

[Clint Byrum]

Post by Theo Schlossnagle

Post by Clint Byrum
mod_backhand is showing erroneous results for the amount of memory on
a machine when running Linux 2.6.5.
Ent Hostname Age Address Total Mem Avail Mem
--- --------- --- ------------- ------------ -----
7 node2 1 1.2.3.5:80 1008 MB 850 MB
11 node3 1 1.2.3.4:80 4294967295 MB 4294967295
node 2 and node3 are identical hardware.. but node3 is running 2.6.5,
and node2 is running 2.4.25. I'm sure this is due to the change in
/proc/meminfo.

Rainer has sent me to patches that should have addressed this problem.
I don't have access to a 2.6.x machine on which I can test
mod_backhand at the moment. Are you running the latest mod_backhand
from CVS?

No.. is there a stable branch that I can trust? And will running CVS in
any way mess up my other servers that are not upgraded yet?

Post by Theo Schlossnagle

Post by Clint Byrum
On a side note, Arriba also comes out *way* lower on this box running
2.6.5. Maybe the box is just slower running 2.6.5 ... or maybe its
the HyperThreading again. :-P

Perhaps... you can manually set the arriba by editing the file it
writes. The arriba is a very arbitrary metric, so candidacy functions
and people should take it with a grain of salt.

Yes definitely. Though I think byLoad might take it too seriously...
because actual traffic across multiple identical machines is directly
proportional to Arriba, even if its only slightly different. I'm not
worried about it... we've started setting arriba manually on identical
machines, and it has worked out well.