Moderators: hgm, Rebel, chrisw
It will likely be a couple of days. Our computer room A/C has once again died, and all three clusters are down, as well as many server machines. We apparently peaked at over 140F in the machine room early this morning, which is scorching hot.
No ETA yet on repairs, we are waiting on them to tell us what has failed.
Let me have the cluster here in Chicago and I can fix the problem easily by opening the windowbob wrote:It will likely be a couple of days. Our computer room A/C has once again died, and all three clusters are down, as well as many server machines. We apparently peaked at over 140F in the machine room early this morning, which is scorching hot.jhaglund wrote:Sent-> craftyje.zipCan you email me a .zip or .tar or whatever (not .rar) of the modified source? I'll run it tomorrow...
I'll look for the results here...
Thanks
No ETA yet on repairs, we are waiting on them to tell us what has failed.
We have one. But nobody lives within 10 miles of the office. The alarm sounded around 5am this morning and started calling people. But with no A/C, and I had the big cluster running at 100%, it gets _hot_ in a hurry. We are hoping nothing was damaged. The tables in the room were actually almost too hot to touch when we got in to power things down.jhaglund wrote:No problem. Funny, you don't have some sort of alarm if the temperature exceeds normal operation levels.
A room temp monitor would be a cheap way to protect an investment.
It will likely be a couple of days. Our computer room A/C has once again died, and all three clusters are down, as well as many server machines. We apparently peaked at over 140F in the machine room early this morning, which is scorching hot.
No ETA yet on repairs, we are waiting on them to tell us what has failed.
Been there in January. And also Minneapolis. Neither is high on my list.michiguel wrote:Let me have the cluster here in Chicago and I can fix the problem easily by opening the windowbob wrote:It will likely be a couple of days. Our computer room A/C has once again died, and all three clusters are down, as well as many server machines. We apparently peaked at over 140F in the machine room early this morning, which is scorching hot.jhaglund wrote:Sent-> craftyje.zipCan you email me a .zip or .tar or whatever (not .rar) of the modified source? I'll run it tomorrow...
I'll look for the results here...
Thanks
Miguel (not a funny joke when you are waiting for the elevated train near the Michigan Lake...)
No ETA yet on repairs, we are waiting on them to tell us what has failed.
As I mentioned, _everything_ is down. Our web server. home directories (I have mine on my own machine, but everyone else uses a departmental file server that is out of service until the A/C is restored. Ditto for clusters, our ftp server, etc. They might power on things like the ftp box as it doesn't produce much heat, but as of noon today everything was dead.Milos wrote:This is a bit offtopic.
Bob what happened to your homepage? It is down since yesterday.
I can only see the sense of this cooling design if everything is designed to keep out the heat of the Alabama summer. For air conditioning you need to have total control over airflow and everything has to be closed. So this could happen with equal frequency if your outside temperatures are hot.bob wrote:We have one. But nobody lives within 10 miles of the office. The alarm sounded around 5am this morning and started calling people. But with no A/C, and I had the big cluster running at 100%, it gets _hot_ in a hurry. We are hoping nothing was damaged. The tables in the room were actually almost too hot to touch when we got in to power things down.jhaglund wrote:No problem. Funny, you don't have some sort of alarm if the temperature exceeds normal operation levels.
A room temp monitor would be a cheap way to protect an investment.
It will likely be a couple of days. Our computer room A/C has once again died, and all three clusters are down, as well as many server machines. We apparently peaked at over 140F in the machine room early this morning, which is scorching hot.
No ETA yet on repairs, we are waiting on them to tell us what has failed.
Northern Minnesota is where I live...We have one. But nobody lives within 10 miles of the office. The alarm sounded around 5am this morning and started calling people. But with no A/C, and I had the big cluster running at 100%, it gets _hot_ in a hurry. We are hoping nothing was damaged. The tables in the room were actually almost too hot to touch when we got in to power things down.
Evaporative cooling will not work here. The typical Summer humidity is in the 60-70% range. It works great in Arizona/New Mexico with their typical humidity of 5-10%. As far as our computer room, it dates back to 1980 or so with respect to design. We have upgraded A/C as we have added machines. And this room is a closed environment with A/C separate from the rest of the building (has to be, the heat load is enormous). There's simply too much heat in too small a space when you look at these clusters that are fairly compact. The cluster I normally use sits in six 6' racks. The heat/density is huge. Even with the A/C running, when I fire up a full cluster run, the air right behind the cluster runs at about 135F. But we have enough A/C and airflow to remove that quickly. Unless one of the four compressors fail. Which happened. Then the maint folks tried to tweak the other three to keep up which led to a complete failure overnight.Eelco de Groot wrote:I can only see the sense of this cooling design if everything is designed to keep out the heat of the Alabama summer. For air conditioning you need to have total control over airflow and everything has to be closed. So this could happen with equal frequency if your outside temperatures are hot.bob wrote:We have one. But nobody lives within 10 miles of the office. The alarm sounded around 5am this morning and started calling people. But with no A/C, and I had the big cluster running at 100%, it gets _hot_ in a hurry. We are hoping nothing was damaged. The tables in the room were actually almost too hot to touch when we got in to power things down.jhaglund wrote:No problem. Funny, you don't have some sort of alarm if the temperature exceeds normal operation levels.
A room temp monitor would be a cheap way to protect an investment.
It will likely be a couple of days. Our computer room A/C has once again died, and all three clusters are down, as well as many server machines. We apparently peaked at over 140F in the machine room early this morning, which is scorching hot.
No ETA yet on repairs, we are waiting on them to tell us what has failed.
But with all the waste heat Bob, of all those machines there must be several things you could do to let that power its own passive cooling? Or would that not work in summer, even in a insulated building, you mentioned you have a very high humidity so that might be too high for cooling by evaporisation etc. But still, the heat should be enough to make some liquid circulation system possible, like the emergency cooling in nuclear reactors, or like in heat pipes? And I would have thought that especially after Katrina there would have been more interest in energy conservation measures, at the time when you bought your clusters at the department I'd guess that was after hurricane Katrina or shortly before Robert?
Regards,
Eelco
A university is supposed to be at the forefront of technology!