Skip to content Skip to main navigation

RiotACT server upgrade underway

johnboy 20 September 2008 64

For the last few weeks we’ve been the victims of our ever-escalating popularity. Our poor server has been spitting out database error messages as often as not, groaning under the increasing weight of all your visits.

In response to this we’ve now lashed out on a dual quad core (so yes eight cores) beast from HP with redundant doodads up the wazoo and 8Gb of RAM (with room for much more). Back when we started in 2000 a machine like this would have come with a team of men in white coats.

We hope it will see us through to complete media domination in Canberra with a start made into the rest of the world.

It’s currently being brought up to speed and we hope to have it up and running in a week or so.

So just stay patient and it will all be better soon. We promise.


What's Your Opinion?


Please login to post your comments, or connect with
64 Responses to RiotACT server upgrade underway
Filter
Order
Danman Danman 1:11 pm 25 Sep 08

I just read this whole thread and my head imploded.

peterh peterh 12:48 pm 25 Sep 08

oh and how goes the upgrade? any news on commissioning date?

peterh peterh 12:48 pm 25 Sep 08

Loquaciousness said :

peter@home said :

Loquaciousness said :

peterh said :

rather the suse geeko.

Teehee. What an apt typo 🙂

L

no typo – they were calling him geeko a while ago. seeing that he is a chameleon, not a gecko…

And so I stand corrected. What an interesting little bit of geek trivia!

L

and now i feel sad that I know it. wait to see the fusa (should have been Foosa) release soon. very interesting. bit like platespin….

Holden Caulfield Holden Caulfield 12:03 pm 25 Sep 08

Any news on the upgrade, farkn???

Jazz Jazz 7:30 am 23 Sep 08

Peter, its on wordpress

Overheard Overheard 12:03 am 23 Sep 08

peterh said :

perhaps if the dl380 g5 was configured with SAS & SATAII drives, you could go with raid 1 + 5 – mirrored O/S and Apps, and raid 5 data store with SATAII – using the WD 1Tb drives, it would give you 4tb actual storage space.

Using that as a virtual ram source, you could also create a HA buffer for ram and allow the max ram config to have a buffer in case of overflow.

SATAII drives would ensure that the spindle speed wouldn’t compromise your DDR Ram bus speed.

Just helped a reseller put together a Promise 14-Bay Raid array, using 1TB SATA drives, 14 bays populated, Raid 5, with FC linkages to a HP server – I think a 560.

Goed ver dommer. Doe je niet in engels praten op het RiotACT? Ik weet het niet wat je hier zegen.

Loquaciousness Loquaciousness 10:03 pm 22 Sep 08

peter@home said :

Loquaciousness said :

peterh said :

rather the suse geeko.

Teehee. What an apt typo 🙂

L

no typo – they were calling him geeko a while ago. seeing that he is a chameleon, not a gecko…

And so I stand corrected. What an interesting little bit of geek trivia!

L

peter@home peter@home 7:47 pm 22 Sep 08

Loquaciousness said :

peterh said :

rather the suse geeko.

Teehee. What an apt typo 🙂

L

no typo – they were calling him geeko a while ago. seeing that he is a chameleon, not a gecko…

Loquaciousness Loquaciousness 7:43 pm 22 Sep 08

peterh said :

rather the suse geeko.

Teehee. What an apt typo 🙂

L

peter@home peter@home 6:37 pm 22 Sep 08

johnboy said :

Peter when a site is being overwhelmed by the number of simultaneous requests a solution that slows down the time to service those requests is more of a problem than a solution.

You’d just increase the backlog and lock it up tighter.

Trust us, we’ve been doing this a while.

out of interest, what is the basis of the riotact environment? is it joomla?

Gungahlin Al Gungahlin Al 5:28 pm 22 Sep 08

Well I’ll just be happy to see it installed, because the errors are giving me the ribbets – especially today…

johnboy johnboy 5:20 pm 22 Sep 08

Peter when a site is being overwhelmed by the number of simultaneous requests a solution that slows down the time to service those requests is more of a problem than a solution.

You’d just increase the backlog and lock it up tighter.

Trust us, we’ve been doing this a while.

Jazz Jazz 4:54 pm 22 Sep 08

Holden Caulfield said :

I’d just settle for a “previous entries” / “next entries” link at the bottom of the page like most other WordPress sites have.
(

Scroll down a bit Holden, It’s already there just labeled with the title of the story rather than next/previous (which is now on the front page too)

peterh peterh 4:54 pm 22 Sep 08

johnboy said :

a) we do not need nor desire a “storage solution”. Storage is not our problem.

b) Virtual RAM? I’d rather shave my balls.

JB, the mental image that conjured up was not required. A simple “no” would have been better.

I have clients (resellers) currently using Virtual ram to enable their servers to bridge temporary out of memory errors. The apps that they are running are sitting on AIX & HP-UX, the HA Virtual RAM is not used continually, it is there as a failover. I wasn’t talking about deploying a swap file, just using VRAM as a backup in case the number of users accelerates yet again. Another Option would be to create several nodes under a Virtual server pool and share the load equally across the nodes. I was assuming that you were using linux, not win for the servers?

johnboy johnboy 4:29 pm 22 Sep 08

a) we do not need nor desire a “storage solution”. Storage is not our problem.

b) Virtual RAM? I’d rather shave my balls.

peterh peterh 3:51 pm 22 Sep 08

caf said :

Kramer: Curious as to why you went with RAID 1 + 0, rather than a small RAID 1 boot partition and big RAID 5 partition for the data? With 4 spindles it should outperform the mirror, and you certainly have the CPU to spare…

peterh: My shot-in-the-dark guess is that the database is hitting its maximum number of simultaneous connections setting, which is largely limited by available RAM.

This is a Hypothetical…
perhaps if the dl380 g5 was configured with SAS & SATAII drives, you could go with raid 1 + 5 – mirrored O/S and Apps, and raid 5 data store with SATAII – using the WD 1Tb drives, it would give you 4tb actual storage space.

Using that as a virtual ram source, you could also create a HA buffer for ram and allow the max ram config to have a buffer in case of overflow.

SATAII drives would ensure that the spindle speed wouldn’t compromise your DDR Ram bus speed.

Just helped a reseller put together a Promise 14-Bay Raid array, using 1TB SATA drives, 14 bays populated, Raid 5, with FC linkages to a HP server – I think a 560. all under $15K for the storage solution. Just because I don’t sell it, doesn’t mean i have forgotten what to do…

johnboy johnboy 3:28 pm 22 Sep 08

Spot on about the RAM caf.

The HD’s are being configured this way because we’re going for availability rather than speed.

We’re nowhere near being limited by HD speed for data access.

caf caf 3:25 pm 22 Sep 08

Kramer: Curious as to why you went with RAID 1 + 0, rather than a small RAID 1 boot partition and big RAID 5 partition for the data? With 4 spindles it should outperform the mirror, and you certainly have the CPU to spare…

peterh: My shot-in-the-dark guess is that the database is hitting its maximum number of simultaneous connections setting, which is largely limited by available RAM.

johnboy johnboy 3:23 pm 22 Sep 08

peterh said :

OT, why does the database keep futzing out? dual 80gb drives should be ok for basic text posts, shouldn’t it?

Maybe back in 1997 with static sites. Here in 2008 with a database driven design it’s all done in RAM with the HD pretty much just a backup for restoring from after a reboot.

tylersmayhem tylersmayhem 3:23 pm 22 Sep 08

Last few days – try last few weeks! I look forward to them stopping after the upgrade.

CBR Tweets

Sign up to our newsletter

Top
Copyright © 2020 Region Group Pty Ltd. All rights reserved.
the-riotact.com | riotact.com.cn | aboutregional.com.au | b2bmagazine.com.au | thisiscanberra.com

Search across the site