From matches at ucc.gu.uwa.edu.au Mon Dec 9 22:03:40 2013 From: matches at ucc.gu.uwa.edu.au (Sam Moore) Date: Mon, 9 Dec 2013 22:03:40 +0800 (WST) Subject: [tech] Mackerel Message-ID: I have recorded Mackerel's system log because that is a fun thing to do on a Monday night. >From what I can gather, Mackerel was a file server (/home ?) from 1992 until 1995. Originally I was going to make a wiki page, but that was effort, so it is in my webspace for now. http://matches.ucc.asn.au/mackerel/small/ Thanks to combto for staying together long enough for me to scan these images, the scanner for bravely scanning regardless of what it saw, image magic for allowing me to produce not quite so ridiculous sized images, and bash for helping me to make the website. Apologies to everyone else. [SZM] From matches at ucc.gu.uwa.edu.au Fri Dec 13 14:36:16 2013 From: matches at ucc.gu.uwa.edu.au (Sam Moore) Date: Fri, 13 Dec 2013 14:36:16 +0800 (WST) Subject: [tech] Curious Message-ID: Curious was hotter than a thousand suns, so I have removed it from the machine room. This affects the progcomp website (which I insist is a UCC service even if only I care about it) and the minutely hostmaster emails about raid. I encourage other people with hot, useless machines to also turn them off for this summer. [SZM] From gozzarda at ucc.gu.uwa.edu.au Fri Dec 13 14:41:19 2013 From: gozzarda at ucc.gu.uwa.edu.au (Andrew Gozzard) Date: Fri, 13 Dec 2013 14:41:19 +0800 (WST) Subject: [tech] Summer Colo shutdown request Message-ID: Hi all, Today, UCC suffered a total network failure as a result of Murasoi overheating. With Summer rolling in, we think it is wise to try and reduce the heat load in the machine room as much as possible. To this end all members with colocated machines in the UCC Machine Room are asked to please remove or power off your machines. While this is not going to be enforced, be warned that UCC services take priority over member colos, and as a result your machine may be shut off without warning. In addition, Musundo has been taken down as a non-critical service that was producing a not insignificant amount of heat. Thanks, [GOZ] From gozzarda at ucc.gu.uwa.edu.au Fri Dec 13 14:44:33 2013 From: gozzarda at ucc.gu.uwa.edu.au (Andrew Gozzard) Date: Fri, 13 Dec 2013 14:44:33 +0800 (WST) Subject: [tech] Additional/upgraded air conditioner Message-ID: Hi all, In the wake of today's overheating issues, it has been proposed that we get and install a better/second air conditioner. Suggestions and comments are welcome. Best Regards, [GOZ] From shmookey at shmookey.net Fri Dec 13 18:24:30 2013 From: shmookey at shmookey.net (Luke Williams) Date: Fri, 13 Dec 2013 18:24:30 +0800 Subject: [tech] Summer Colo shutdown request In-Reply-To: References: Message-ID: I use my colocated server (sophocles) extensively and the service is very valuable to me. Is this the kind of problem that can be solved with money? The alternative of commercial colocation is really expensive (>$1000/year) so I wouldn't feel hard done by if the club asked for a substantial fraction of that as a regular fee for members to keep their colos. If the cost deters enough low-use hardware to obviate the immediate heat issue, perhaps the money could be put into a special fund for improving the machine room. All under the broad protections of our standard SLA, of course. Thoughts? On Fri, Dec 13, 2013 at 2:41 PM, Andrew Gozzard wrote: > Hi all, > > Today, UCC suffered a total network failure as a result of Murasoi > overheating. With Summer rolling in, we think it is wise to try and reduce > the heat load in the machine room as much as possible. > > To this end all members with colocated machines in the UCC > Machine Room are asked to please remove or power off your machines. > > While this is not going to be enforced, be warned that UCC services take > priority over member colos, and as a result your machine may be shut off > without warning. > > In addition, Musundo has been taken down as a non-critical service that > was producing a not insignificant amount of heat. > > Thanks, > [GOZ] > Unsubscribe here: http://lists.ucc.gu.uwa.edu.au/mailman/options/tech/shmookey%40shmookey.net From bob at ucc.gu.uwa.edu.au Fri Dec 13 20:15:07 2013 From: bob at ucc.gu.uwa.edu.au (Andrew Adamson) Date: Fri, 13 Dec 2013 20:15:07 +0800 (WST) Subject: [tech] Summer Colo shutdown request In-Reply-To: References: Message-ID: As I understand it, you're not using sophocles for any hardware experimentation, so is there any reason it can't be turned into a VM on medico? It's just that sophocles uses the old P4 generation xeons which run really hot. If you really want your own physical machine that has to be on all the time, could you upgrade to something that is substantially cooler? Andrew Adamson bob at ucc.asn.au |"If you can't beat them, join them, and then beat them." | | ---Peter's Laws | On Fri, 13 Dec 2013, Luke Williams wrote: > I use my colocated server (sophocles) extensively and the service is > very valuable to me. Is this the kind of problem that can be solved > with money? > > The alternative of commercial colocation is really expensive > (>$1000/year) so I wouldn't feel hard done by if the club asked for a > substantial fraction of that as a regular fee for members to keep > their colos. If the cost deters enough low-use hardware to obviate the > immediate heat issue, perhaps the money could be put into a special > fund for improving the machine room. All under the broad protections > of our standard SLA, of course. > > Thoughts? > > > On Fri, Dec 13, 2013 at 2:41 PM, Andrew Gozzard > wrote: > > Hi all, > > > > Today, UCC suffered a total network failure as a result of Murasoi > > overheating. With Summer rolling in, we think it is wise to try and reduce > > the heat load in the machine room as much as possible. > > > > To this end all members with colocated machines in the UCC > > Machine Room are asked to please remove or power off your machines. > > > > While this is not going to be enforced, be warned that UCC services take > > priority over member colos, and as a result your machine may be shut off > > without warning. > > > > In addition, Musundo has been taken down as a non-critical service that > > was producing a not insignificant amount of heat. > > > > Thanks, > > [GOZ] > > Unsubscribe here: http://lists.ucc.gu.uwa.edu.au/mailman/options/tech/shmookey%40shmookey.net > Unsubscribe here: http://lists.ucc.gu.uwa.edu.au/mailman/options/tech/bob%40ucc.gu.uwa.edu.au > From bob at ucc.gu.uwa.edu.au Fri Dec 13 21:06:26 2013 From: bob at ucc.gu.uwa.edu.au (Andrew Adamson) Date: Fri, 13 Dec 2013 21:06:26 +0800 (WST) Subject: [tech] Additional/upgraded air conditioner In-Reply-To: References: Message-ID: The problem with the current cooling is twofold: 1. We've added several machines to the machine room since the aircon was specced and installed 2. The aircon we have isn't being used to its full potential due to the airflow in the room The first issue will be largely solved in the near future with the new storage server and the retirement of the SAN. This will remove the thermal load of several shelves of disks as well as two P4 era servers. There are other things we can look in terms of hardware, such as virtualising some of the older hardware machines (progcomp, pomona, sophocles, dagon), but I suspect the new storage server will solve a lot of our problems. The second issue can again be split into two-issues; the cool air isn't being evenly directed towards the three racks (causing the right rack to be hotter), and the hot air being generated by all three racks isn't being extracted properly. Both of these issues can be resolved with some blanking plates on the racks, an air baffle on the aircon to direct the air, and if necessary it's relatively easy to design and build a duct with a blower on it to extract the hot air from behind the racks to outside. On a side note, we have the option on the aircon of taking uncooled air from either inside or outside. At the moment the source of air is selected manually using a lever on the front of the aircon. Would it be worth making something that can automate this action in order to be cooling the air that is already the coolest at any one time? Andrew Adamson bob at ucc.asn.au |"If you can't beat them, join them, and then beat them." | | ---Peter's Laws | On Fri, 13 Dec 2013, Andrew Gozzard wrote: > Hi all, > > In the wake of today's overheating issues, it has been proposed that we > get and install a better/second air conditioner. Suggestions and comments > are welcome. > > Best Regards, > [GOZ] > Unsubscribe here: http://lists.ucc.gu.uwa.edu.au/mailman/options/tech/bob%40ucc.gu.uwa.edu.au > From bob_george33 at hotmail.com Sat Dec 14 15:39:13 2013 From: bob_george33 at hotmail.com (bob_george33) Date: Sat, 14 Dec 2013 15:39:13 +0800 Subject: [tech] [wheel] Summer Colo shutdown request Message-ID: I am for this. And then rebuilding it. I will spec it up in the next week. Bobgeorge33 Sent from Samsung Mobile -------- Original message -------- From: Anil Sharma Date: 2013/12/14 04:12 (GMT+08:00) To: Bob Adamson Cc: wheel ,Luke Williams ,tech at ucc.asn.au Subject: Re: [wheel] [tech] Summer Colo shutdown request TEAR THE MACHINE ROOM DOWN On 13/12/2013 1:15 PM, "Andrew Adamson" wrote: As I understand it, you're not using sophocles for any hardware experimentation, so is there any reason it can't be turned into a VM on medico? It's just that sophocles uses the old P4 generation xeons which run really hot. If you really want your own physical machine that has to be on all the time, could you upgrade to something that is substantially cooler? Andrew Adamson bob at ucc.asn.au |"If you can't beat them, join them, and then beat them." ? ? ? ? ? ? ? ?| | ---Peter's Laws ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?| On Fri, 13 Dec 2013, Luke Williams wrote: > I use my colocated server (sophocles) extensively and the service is > very valuable to me. Is this the kind of problem that can be solved > with money? > > The alternative of commercial colocation is really expensive > (>$1000/year) so I wouldn't feel hard done by if the club asked for a > substantial fraction of that as a regular fee for members to keep > their colos. If the cost deters enough low-use hardware to obviate the > immediate heat issue, perhaps the money could be put into a special > fund for improving the machine room. All under the broad protections > of our standard SLA, of course. > > Thoughts? > > > On Fri, Dec 13, 2013 at 2:41 PM, Andrew Gozzard > wrote: > > Hi all, > > > > Today, UCC suffered a total network failure as a result of Murasoi > > overheating. With Summer rolling in, we think it is wise to try and reduce > > the heat load in the machine room as much as possible. > > > > To this end all members with colocated machines in the UCC > > Machine Room are asked to please remove or power off your machines. > > > > While this is not going to be enforced, be warned that UCC services take > > priority over member colos, and as a result your machine may be shut off > > without warning. > > > > In addition, Musundo has been taken down as a non-critical service that > > was producing a not insignificant amount of heat. > > > > Thanks, > > [GOZ] > > Unsubscribe here: http://lists.ucc.gu.uwa.edu.au/mailman/options/tech/shmookey%40shmookey.net > Unsubscribe here: http://lists.ucc.gu.uwa.edu.au/mailman/options/tech/bob%40ucc.gu.uwa.edu.au > Unsubscribe here: http://lists.ucc.gu.uwa.edu.au/mailman/options/tech/maset%40ucc.gu.uwa.edu.au -------------- next part -------------- An HTML attachment was scrubbed... URL: http://lists.ucc.gu.uwa.edu.au/pipermail/tech/attachments/20131214/a1c7dd0a/attachment.htm From nick at ucc.gu.uwa.edu.au Sat Dec 14 19:58:58 2013 From: nick at ucc.gu.uwa.edu.au (Nick Bannon) Date: Sat, 14 Dec 2013 19:58:58 +0800 Subject: [tech] antarctica, vs Additional/upgraded air conditioner In-Reply-To: References: Message-ID: <20131214115858.GV2270@ucc.gu.uwa.edu.au> On Fri, Dec 13, 2013 at 09:06:26PM +0800, Andrew Adamson wrote: [...] > 1. We've added several machines to the machine room since the aircon was > specced and installed > 2. The aircon we have isn't being used to its full potential due to the > airflow in the room 3. We already have all the air conditioning we can cope with. We're (still!) not paying for power, but we can't suck unlimited amounts. [BOB] organised getting antarctica, the $1100? Kelvinator, around 2010-04-09 - it's proven itself as a pretty good buy, heading into its fourth summer now. 5.3KW of cooling, about 10A/2400W consumption. Previously we had one, then two, small aircons instead of one large one: simpson/gobi/gibson/atacama/judean. In case of a single failure, two small ones does buy us a few extra minutes before temperatures hit critical, if that helps. An aircon is a consumable for us, gibson made it through one summer, started icing up and being unreliable in its second year (causing machine room outages) and died. [DAA] got it fixed under warranty after much chasing, but it didn't last to another summer. On a 42 degree day, 5.3KW of cooling (in as-new condition, with good airflow) means that we can have about 5.3KW of servers running. If the aircon fails or its power circuit trips, temperatures will go critical in a few minutes, probably before anyone is around to react usefully. If antarctica starts icing up or failing, we're not going to do much planning/fiddling/approving at that point, we'll replace it. Chances are that it'll happen this time of year-ish; and chances are that the aircon will still be in warranty, but hard to prove that it's irreparably broken. If/when it is fixed, it just means we have a spare. [...] > On a side note, we have the option on the aircon of taking uncooled air > from either inside or outside. At the moment the source of air is > selected manually using a lever on the front of the aircon. Would it be > worth making something that can automate this action in order to be > cooling the air that is already the coolest at any one time? Handy in an emergency... but on a warm day, we expect the machine room to be below ambient temperature. If it's not, we've got too much turned on. Nick. -- Nick Bannon | "I made this letter longer than usual because nick-sig at rcpt.to | I lack the time to make it shorter." - Pascal