| |
| |
| |
|
Page: 1 2 3 4
Comments:
<0> milo, hi
<1> while you're at it
<2> Invalid inr unable to find computer add^H^H^H^H^
<2> haha
<2> hmm
<1> milo: no
<1> the drug channels are further down the corridor
<0> where the fancy lights are ;)
<1> aXp: you'd know :P
<0> i don't use drugs anymore!
<0> ;)
<2> http://pastebin.com/723045
<0> they told me to stop it ;_
<1> aXp: I bet you do :P
<2> that is the output of cat /dev/tts/USB0
<2> interesting isnt it
<1> not really :)
<2> stfu
<2> pretend
<1> no acronyms please :P
<0> batphone, just minicom running?
<1> if you cant be nice then /part :)
<2> yes
<2> you be nice
<2> aXp when i stop minicom it still scrolls for a couple minutes
<3> someone know if kernel 2.4.27 has a bug in fs/inode.c?
<0> no cat or similar? dialing turned off in minicom?
<0> milo, sure on line 721!
<2> minocom -o
<0> yep!
<3> aXp: I could see this bug in 1238 line.
<3> kernel BUG at inode.c:1238
<1> is this with your kernel or a dist based one?
<3> gb: debian dist
<3> [vfs_readdir+147/224] [nfs3svc_encode_entry_plus+0/48]
<3> and
<1> ahh ok, good news :)
<0> maybe you pressed some fancy key-combination to start zmodem or something else which sends weired sequences! which terminal emulation do you use! try ANSI,VT100,VT102
<3> [vfs_unlink+329/560] [nfsd_unlink+254/496]
<0> milo, it doesn't necessary mean that there is an error in the kernel, maybe yout inode tables are ****ed up! or disk problem
<3> gb: should I change my debian kernel for printine?
<0> printine?
<3> aXp: sorry
<1> milo: I was actually asking batphone
<3> pristine
<0> pristine?
<3> oficial kernel
<1> milo: I suggest booting into a stable kernel & doing an fsck
<0> or try to get some S.M.A.R.T info of your drive, although don't believe everything SMART tells you!
<1> lots of things can cause nfs related errors & hence corrupt file systems
<1> a very good reason too mount nfs read only if you need too access information that doesn't need writing too
<1> or too use other less sensitive protocols for writing too files across a network
<3> aXp: I am using theerteen xfs filesystem on storage EMC (raid5) with rw access for ninety clients
<3> so I dont think that problem can be disk or fs
<1> milo: thats the problem with asking for advice without displaying all the facts up front
<0> milo, k ;)
<1> & it still could be disk or fs
<1> I've seen raids collapse before & filesystems similar self destruct
<3> gb: what the kernel I should to use for one stable nfs server + xfs?
<1> milo: theres no simple answer
<3> gb: :(
<1> its a **** it & see type od scenario by the sounds of it till you get it right
<3> gb: ok
<1> access too that server must be painfully slow though
<1> nfs aint exactly the fastest network filesystem protocol
<1> why do you have everything on one server?
<3> gb: I have more of ninety diskless servers in production mode...
<1> eww, thought diskless was a thing of the past
<4> can someone explain this to me? --> http://developer.novell.com/wiki/index.php/Category:Partner_Linux_Driver_Process
<3> gb: not for us
<1> Can?Perhaps;Want too?Perhaps not :)
<3> gb: 2.6 series is more stable for this job?
<1> milo: no idea, I've not had too mount so many boxes off one server for nfs, at most I'd only do 10boxes with nfs because of it's slowness
<3> gb: ok tks
<1> my guess is the congestion is the cause of your errors
<3> gb: no gb, my network is 1000mbit and disks are in EMC(fast io)
<3> I have many servers in this case: smtp servers, http, databases, etc
<1> milo: makes no difference, you still have the bottleneck of one box handling all your i/o
<3> offcourse, those that need IO I put one hba
<0> milo, it would be best to try another kernel!
<3> aXp: what?
<1> if that kernel is struggling & giving you errors, by all means try something else
<0> how long have you used your current kernel? did you expierence this problem often?
<1> I personally think the kernel is probably not the cause of the problem
<3> aXp: one time for week
<1> milo: I think you misintepret what I mean by servers, by servers I mean actual computers holding the serving raid
<1> one computer handling 90 requests simultaneously for the same data has too process all the gigabyte delivered requests, read the emc disks & deliver it back over the same gigabyte network
<3> gb: this enviroment still working for 2 years
<1> milo: I dont doubt it works, but I'm saying the errors are probably there for good reason
<1> if they've only started after a kernek upgrade then rollback or try a later kernel
<1> arrgh
<3> gb: this is one conceptual error...
<3> gb: I am talking of my enviroment...
<3> ;-)
<3> gb: sincerely I am thinking in put one solaris box for this job.
<3> well SUN makes this protocol.
<0> NFS ?
<3> yes nfs
<0> v2,3,4?
<5> Is there atheros chipset drivers in the current kernel...they seem to have disapeared.
<0> there nevere was one
<0> madwifi[ng]
<5> aXp: yeah, that one
<5> hmm...strange so where can I get madwifi
<3> well tks axp and gb for yours patience.
<0> google
<1> milo: when I was at uni last millenium we had half a dozen suns & sg's nfs together too do a similar thing for hundreds of terminals, but they were all vt50 style terms & there were several boxes for handling the i/o & they struggled
<5> aXp: so madwifi is not in the kernel, where do I get it...
<0> http://madwifi.org/
<5> aXp: wow thanks...appreciated
<0> they are well documented! look at the newbie section!
<0> you are welcome
<1> milo: in short centralised anything has always been a bad idea & leads too bottlenecks & performance issues
<2> whats a good way to have hot swappable drives in a raid array so that you can rebuild an array while the machine is up after replacing a drive?
<0> batphone, hw raid
<2> i need to build a 2 TB storage array
<2> scsi is probably going to be the best way to go
<6> ??
<0> but expensive!
<6> maybe
<0> SCA
<2> it has to be as reliable as possible
<6> i have a small scsi setup
<6> its nice
<2> http://www.millertech.com/indpc/servers/RAID/8_Bay_SCA_RAID.htm
<2> this looks cool
<2> how would you swap a drive and rebuild a hardware raid array onto it while the machine is up?
<1> Why do you say scsi is the best?
<2> speed and reliability... as opposed to SATA w/mdraid
<1> it's not best in terms of cost even speed/reliability can be comparable too modern ide
<6> gb, ??
<6> the value is lower
<0> but hw-solutions are still better then any md raid
<6> but scsi drivers are still of way more solid build quality
<1> I prefer scsi but I've seen just as many dead scsi disks as I have ide
<0> or sata raids
<6> gb, maybe you should considder the usage pattern of scsi vs ide
<1> hw raid is always preferred over software raid
<6> most ide drives just sit and relax all day
<0> i've seen many ide disks dying, but my good ol 9,1GB IBM 10k u2w scsi still works like a charm
<6> yea
<1> chn: maybe I'll just rely on my past experiences
<2> i worked in a datacenter with over 8000 linux servers for 2 years
<2> IDE ****S
<2> sata is not that much better
<6> why should it
<6> same **** different interface
Return to
#kernel
or
Go to some related
logs:
#web
#debian
group by variable restraints sql
Gaim IRC Helper
USE=opengl gentoo kdenetwork
sshfs fusermount: mount failed: Invalid argument
#fedora
x86 40x80
gnomebaker Wrong medium type
config dual head nv xorg dfp
|
|