QuicksearchNavigationEventsTrusted AdBlower Door Test
Luftdichtigkeitsmessung für Gebäude Raum Hamburg ab 160 € excl. MwSt www.m-tectum.de Kategorien
|
RantsTuesday, April 17. 2007Trackbacks
Trackback specific URI for this entry
No Trackbacks
Comments
Display comments as
(Linear | Threaded)
Definitely. The M2 is a massive failure. Crappy chipsets and broken firmwares.
I prefer some more elaborated criticism. What are the points of your critic ?
The used nvidia chipset is imho a desktop chipset and not one for servers. It was a serious pain to patch linux against error messages arising from the forcedeth nic driver. The firmware 1.11 is not useable (video redirection is broken) when you have a german localisation on your computer. This is even worse when you get a bunch of M2 and have to rollback the firmware.
After this much of trouble and work the galaxy without the M2 is the better choice.
Regarding the chipset: I don´t see the problem of using an nforce professional chipset. It´s the NVIDIA offering for server and workstations. IBM use ist, HP use it. And although it´s derived from a desktop chipset, it was modified for server use.
Regarding network device : Can you offer more data about the linux/network driver stuff ? Do you have a case id for the ILOM problem. Do you have the same problem with newer versions ?
Bugs: 6513809, 6527969, 6494290, 6505164 from here. The word 'sometimes' in 6513809 is great understatement - not one of a dozen were working. For weeks now there is no new version. An hotfix exists but iirc this has other implications.
Regarding the fordeth driver there were errors like 'too many iterations in nv_nic_irq' which lead to numerous kernel patching and building. Problems which we didn't had with normal Galaxies.
Bugs: I´ve talked to a specialist, the problem is fixed. But i will check it, when i´m back in hamburg
forcedeth: As far as i know the problem was fixed in recent kernels. But well, this is a problem you have with development in Linux ...
Is this a response?
http://blogs.sun.com/jonathan/entry/what_service_means_to_me or just a strange coincidence?
I have no informations right now, but at the moment i would say: Strange coincidence
Like Ben writes, I also don't like the speed of docs.sun.com. The documentation itself is great, but searching is a PITA...
The same goes for searching on sunsolve. It's much easier to open a support case than trying to find a patch for a problem. Maybe Sun uses internally another search engine? What disturbs me the most about LOMs is, there are so many different. Why are they different on new Opteron systems, and even on sparc you have different types. At least the interface could be similar... Ever tried to log out of a LOM? Was the command for it quit, exit, bye, logout or logoff or maybe ctrl-d?
Sun uses the same SunSolve technology internally - just filled with a little bit more information
When I worked for Sun i always complained about the speed of SunSolve. Now beeing a customer, I wish I could even the speed of internal SunSolve... Usually, about 25% of my logins to SunSolve fail because of timeouts (Slow auth server?). WTF??? I also can confirm the problems with OSC. It's plain slow and crappy. Customers aare complaining about speed and usability since years but nothing happens.
Regarding Ben's rant:
1. SunSolve No, it's not "utterly useless". Even for a rant, that's just not fair. Actually, the quality and depth of documentation overall is one of the reasons why I tend to prefer Sun systems over other 1st tier vendors. Speed and also availability of the SunSolve website, however, is really awful. In addition, for a company like Sun that advertises expertise in high availability and throughput, one has to wonder: either they can't get it done, or they don't care. Let me state clearly that "slow" and "availability" doesn't mean I feel it could be a little faster or that I dislike a downtime once a week. What I mean is: about 1 in 5 times I use SunSolve, it is either completely down or so slow that login times out (several times in sequence). At the same time, the regular Sun website (Germany and US) is up and performing well. Definitely not a connectivity problem here... 2. X4200 So far I didn't even notice that Sun has downgraded 2 of the 4 nics in the M2 series. Can't say I like that. Beginning with the fact that I don't like supporting multiple network drivers in one machine. This sucks. Do that in the X2** series if you need to. No compromise in the X4** please. Regarding Ben's praise for the original X4100/4200: I mostly agree, however the choice of the LSI logic controller was a poor one. We had several problems including a broken RAID 1 on the local disks (unable to boot) that Sun Support was unable to solve except by offering a new mainboard. In the end we did it ourselves. We just decided that we don't trust LSI logic controllers, removed power completely, resetted everything to factory defaults, then reinserted the disks and asked the controller to rebuild. Worked and still does. As for firmware issues: the firmware in several of the current X86 Sun systems is less that perfect. Upgrade sequences for BIOS, LSI-BIOS, iLOM are a major PITA. Take the bare X4200, insert a Sun rebranded SCSI controller (turns out to be LSI again), also insert a current enterprise class Emulex fibre card. Turns out you have to disable the Emulex BIOS because... the memory reserved for expansion cards is already exceeded). Great to know why we ordered a machine with 6 expansion slots. So.. The original X4200 is a brilliant machine, except for the LSI stuff and except for the firmware pains.
mhh, interessting we are just about to buy some X4200 M2 servers. We cannot use a different system, because we want do to use Xen with Pacifica. Hopefully we are going to get a test system. If we need to use only 2 nics, do you think i should give it a chance?
Maybe Sun should push Nvidia to open their chip specs to open source developer. I consider nge troubles because Nividia did not open their specs for open source developer, and not because their chip design is just crap, right?
There is nothing inherently bad of the nforce ethernet nics.
1. The nge driver in Solaris ist supported as any driver in Solaris. They are fully supported by Sun. 2. Even the linux drivers are not bad. 3. When you only need 2 interface and you don´t trust the Ethernet you can still use only two e1000 interfaces. I see no problem in using it ...
> Hopefully we are going to get a test system.
Sure, why not? Just get a try and buy system directly from Sun. Usually takes about 1-2 weeks. > If we need to use only 2 nics, do you think i should give it a chance? As far as I can see (no personal experience with the M2 version), it's still a brilliant machine. Just make sure you try everything out before going productive. Get the latest firmware up and running. Install the OS you want/need and stress your system to make sure drivers work ok. I would do that with any system, however. If you use the Raid 1, take a disk out and check if everything works. If you need expansion cards, check that the option rom space suffices. If you need to boot from an external device, check the Release Notes, there are some restrictions regarding what work from which slot. You may need some iterations until everything is fine, but when it is, it stays that way. > I consider nge troubles because Nividia did not open their specs for open source developer, and not because their chip design is just crap, right? I don't like device vendors that hide/nondisclosure interface specs. This is bad engineering ethics and practice, IMO. As for the NGEs: I can't really comment on how good or bad they are. Still, I personally hate supporting multiple drivers in a machine just for the fun of it. If you don't use the NGEs, no problem there.
A side note The problems with the nforce ethernet dates back to kernel from summer last year as far as i remember.
The described problem was a problem of Linux, not of the system. But i understand the comment of Joern: The e1000 has earned the reputation of beeing the network driver with the highest quality throughout all open source operating systems. |
The LKSF bookThe book with the consolidated Less known Solaris Tutorials is available for download here
Web 2.0Contact
Networking open.bc My photos SyndicationComments about INTJ Thu, 04.12.2008 23:55 "Detected language German. The only supported languages are English and Swedish. But stay tuned, we are expanding. [...] about Leaked Airbus "787 lessons learned" presentation Thu, 04.12.2008 22:09 Funny .. i guess another lesso ns learnt case would be: Do not circulate too much confid ential information among [...] about Overpopulated Thu, 04.12.2008 17:36 Oh yeah, try being on a dual q uad core and running MenuMeter s, just the CPUs alone take up a lot of space. Can [...] about Deutschsprachiger Anwenderbericht zu Opensolaris 2008.11 Thu, 04.12.2008 15:41 Jo ... einige Dinge sind nicht ganz korrekt ... aber im gros sen und ganzen finde ich den T ext gut about Deutschsprachiger Anwenderbericht zu Opensolaris 2008.11 Thu, 04.12.2008 15:17 "Ab Version 2008.11 kann man a uch von ZFS booten." na das ging aber auch schon mit 2008 .05... Getaggte ArtikelAMD Apple avs Bahn Blogging Blogosphere braindump Business Travel CeBIT cec cec2006 CMT del.icio.us deutsch dtrace fliegen Fundsache General Hamburg IBM i hate sundays Intel iscsi jumpstart Links Linux lksf Mindfuck Movies Music Musik Niagara Opensolaris Opteron Photographie policy of ... Politik Security Solaris storage Sun suncec2007 sunw t1 The IT Business Ultrasparc ultrasparc t1 Wirtschaft Work ZFS
Blog Administration |