Web server upgrade, Saturday 11/1

[linp]

I will be going to the server room tonight around 8 or 9PM to add another 512MB RAM to our Web server. It *should* be down for about 15 mins.

Thank you so much for the donation and support!

[GMAN]

Mmmm...RAM good. BBS performance after the upgrade should shoot through the roof, right? :D

[linp]

RAM added, 1 friggin gig. Now I can tinker with the web/database server settings hehe.

Notice any difference in speed, and responsiveness, database errors while lots of people are on?

[dnakase]

Uh, like what? What should I notice?

[GMAN]

Do you have any 'before' dumps of vmstat or top or equivalent? How many I/O writes have we been averaging?

I think you can start messing with the database query caching now. :)

Oh, was the system pegging the CPU, memory, or pagefile before the RAM upgrade?

[linp]

I don't have any "before" dumps of top/vmstat.

CPU is plenty fast, rarely ever see the load average go much higher than 2, and usually >50% idle.

Memory was getting exhausted at times, and swap gets used to a large amount. I remember the system always having less than 100MB of free "real" ram with just a moderate amount of activity. Now it's showing about 400-500MB free. Hopefully swap will never be touched.

Current:
<code style="white-space: nowrap">last pid: 1083; load averages: 0.13, 0.18, 0.15 up 0+00:58:20 21:44:36
48 processes: 1 running, 47 sleeping
CPU states: 0.4% user, 0.0% nice, 0.4% system, 0.4% interrupt, 98.8% idle
Mem: 253M Active, 226M Inact, 74M Wired, 18M Cache, 112M Buf, 434M Free
Swap: 1024M Total, 1024M Free</code>

<code style="white-space: nowrap"> procs memory page disks faults cpu
r b w avm fre flt re pi po fr sr ad0 ad2 in sy cs us sy id
1 0 0 535240 479284 285 2 1 0 247 0 0 0 312 3354 206 10 2 88</code>

[GMAN]

That's cool that the CPU doesn't get pegged. Looks like the RAM upgrade should last for a while.

What tweaks do you have planned for mySQL?

[linp]

I think my mysqld settings are ok. Gonna fiddle with # servers in Apache settings.

[GMAN]

Lemme know if you need a little HTTP load-testing done.

800,900,1000,10000 requests per second...it's your call. All of this will be coming from a T3 pipe.

:D

[idukes]

I can't believe the SY reading is 3354, that's unheard-of! :D

Ian

[FuzyTech]

Now all's you have to do is re-rasterize your pixel port mapping buffer and all will be fine....

...Damn computer geeks! :D

[blk650]

fuzy, you forgot to mention letting the crystal template re-boot so the initialization of the flux capasitor will properly re-initiate during peak usage times.


but really, in case you thought i was serious, i have no idea what you do but keep doing it. thanks guys.

[ZombieDude]

Originally posted by linp

Current:
<code style="white-space: nowrap">last pid: 1083; load averages: 0.13, 0.18, 0.15 up 0+00:58:20 21:44:36
48 processes: 1 running, 47 sleeping
CPU states: 0.4% user, 0.0% nice, 0.4% system, 0.4% interrupt, 98.8% idle
Mem: 253M Active, 226M Inact, 74M Wired, 18M Cache, 112M Buf, 434M Free
Swap: 1024M Total, 1024M Free</code>

<code style="white-space: nowrap"> procs memory page disks faults cpu
r b w avm fre flt re pi po fr sr ad0 ad2 in sy cs us sy id
1 0 0 535240 479284 285 2 1 0 247 0 0 0 312 3354 206 10 2 88</code>

I once met a guy who said he knew another guy who cousin had a girlfriend that lived in the Niagara Falls area. This girlfriend's dad had best friend who's neighbor had a 68 Camaro with a Holly Double Pumper Hemi, running open headers that had performance similar this except he had load averages in the 0.22, 0.21, 0.11 range.
Strange...

[linp]

Strange indeed... after the box has warmed up, these are the readings, more in line with her 68 Camaro:
<code style="white-space: nowrap">last pid: 32908; load averages: 0.14, 0.20, 0.15 up 1+01:50:45 22:37:01
57 processes: 2 running, 55 sleeping
CPU states: 3.3% user, 0.0% nice, 0.0% system, 0.0% interrupt, 96.7% idle
Mem: 487M Active, 185M Inact, 102M Wired, 47M Cache, 112M Buf, 183M Free
Swap: 1024M Total, 84K Used, 1024M Free
</code>
<code style="white-space: nowrap"> procs memory page disks faults cpu
r b w avm fre flt re pi po fr sr ad0 ad2 in sy cs us sy id
1 0 0 853572 227296 315 1 0 0 248 11 0 0 306 4675 317 11 3 86
</code>

[Punkbrad]

does this mean computer stuff is happening?

website - good

no website - bad

Pb.

[Reddog99]

With that much idle power, you should be running some distributed computing project like Folding@Home. Then you get stats like these:

Processes: 63 total, 4 running, 59 sleeping... 171 threads 00:58:30
Load Avg: 1.51, 1.55, 1.38 CPU usage: 51.9% user, 48.1% sys, 0.0% idle
SharedLibs: num = 54, resident = 18.7M code, 1.71M data, 5.73M LinkEdit
MemRegions: num = 4704, resident = 67.4M + 13.5M private, 84.9M shared
PhysMem: 54.5M wired, 113M active, 198M inactive, 366M used, 274M free
VM: 2.20G + 37.5M 18920(0) pageins, 0(0) pageouts




Pat

[linp]

I will be adding another 512MB RAM to our web server around noon TODAY. So, expect a few minutes of downtime.

[bwarbiany]

Can we get back on a motorcycle-related topic please?! ;)

J/K Paul, you're doing a great job

[Golden_Eternity]

Posted to the CB by shady adam -

shady adam: server should be up for a while now.. was upgrading some stuff (admin)
shady adam: Will be down for like 5 minutes again in an hour or two but that's it

[adam]

Done with all the updates.

-a

[No_Brakes23]

add another 512MB RAM to our Web server.

Thanks, Paul.

Now I feel even guiltier for not donating yet.