No difference between servers yet different performance.

[das329717]

I am quite stunned and thay why I'm posting here. For almost a year now, I'm running 2 EG AMD's from OVH... Both got a 6-core AMD-CPU and 16GB RAM and were manually installed with software RAID-0 and Debian 6. On top of that, manual install of rtorrent and rutorrent, basically compiled from the various tutorials on the internet. Got a few cron-jobs running to update the kernel where needed, clean up logs and such but no big deal.

For some reason, unknown to me, one of the machines is recently acting up under load by which I mean having more than 200 loaded torrents, maybe 30-40 transferring data and 1-3 torrents downloading. Where machineA will easily do 50MB/s incoming and 30MB/s outgoing, machineB will either accept high incoming OR high outgoing but not both at the same time. And when I use 'top' to check the memory, machineB has got a tremendous amount of 'cached' in use where machineA hasn't. Tried recompiling the entire rtorrent-installation but to no vain. Tried diminishing the rtorrent.rc but also no use. Booted in rescue-mode and no errors found. So, not to compromise the current installs, I executed my own install.sh on a virtual machine at home. Although I can only test with smaller amounts of active torrents, the install seems allright since it doesnt start filling up 'cached' here as well.

Quote

top - 10:46:51 up 23 min, 2 users, load average: 0.31, 0.30, 0.25
Tasks: 119 total, 1 running, 118 sleeping, 0 stopped, 0 zombie
Cpu(s): 2.6%us, 1.2%sy, 0.0%ni, 94.2%id, 1.4%wa, 0.0%hi, 0.6%si, 0.0%st
Mem: 16460764k total, 11077088k used, 5383676k free, 7612k buffers
Swap: 61437936k total, 0k used, 61437936k free, 10491984k cached​

Can anyone give me some pointers where to look or, if needed, I can post some config-files from the machine to help troubleshoot the system.

 

[jith45]

Get IOSTAT and check the CPU wait time..

How many/how fast are your drives?

Seeing that you have 200 torrents the first thing that comes to mind is how the OS/filesystem handles all of the open sockets.
This statement "machineB will either accept high incoming OR high outgoing but not both at the same time" points a finger at OS resource management.
Is machineB on it's own separate internet connection? It's also possible it's an internet problem, your LA's look fine

 

[das329717]

Code:
avg-cpu: %user %nice %system %iowait %steal %idle
0.56 0.00 1.62 19.39 0.00 78.44

Device: tps Blk_read/s Blk_wrtn/s Blk_read Blk_wrtn
sda 168.00 3174.40 139683.20 15872 698416
sdb 162.40 3139.20 137929.60 15696 689648
md6 28937.20 5840.00 231115.20 29200 1155576
md1 1.20 0.00 9.60 0 48
Seconds later, still while downloading @ 60MB/s
Code:
avg-cpu: %user %nice %system %iowait %steal %idle
5.48 0.00 6.35 2.87 0.00 85.30

Device: tps Blk_read/s Blk_wrtn/s Blk_read Blk_wrtn
sda 33.80 16353.60 1280.00 81768 6400
sdb 30.40 16464.00 1272.00 82320 6360
md6 582.60 32817.60 2475.20 164088 12376
md1 4.40 0.00 35.20 0 176
Drives are, from what I know, 7200RPM SATA 2TB. Two of those are mounted in an EXT4 RAID0-set with 95% allocated to md6. Yet, when I do the same download (same torrent, same RuTorrent-setup and such) on machineA, iostat won't exceed 5%.

Don't know if I clearly stated it in the first post but both servers were installed with the exact same script, have had the exact same updates, patches, configs and such. One behaves nicely, other one displays this behavior.

 

[jith45]

you could have a server with a hardware issue or on a different network segment. I've had the exact same issue on ovh before. They are a budget hosting company for a reason. They are also famous for giving different hardware than what they advertise.
Also ovh does tend to traffic shape some servers.

 

[das329717]

you could have a server with a hardware issue or on a different network segment. I've had the exact same issue on ovh before. They are a budget hosting company for a reason. They are also famous for giving different hardware than what they advertise.
Also ovh does tend to traffic shape some servers.

I kinda know your opinion on OVH from other posts. And yeh, I agree on the hosting-part although so far I had not had any issues with getting other/inferior hardware from them. Backtracked my servers; according to their weathermap, the machines are in the same location and only 2 racks apart. Speeds are good though but not in & outgoing at the same time. Hence I do think there might either be a problem with the server-hardware or the OS.

Is there any way I can test/report the disks from within Debian to see if there are issues with the disks, partition-table, broken blocks and such without "breaking" the RAID-functionality?