[NTLUG:Discuss] writing to a USB drive locks up server..

[LEROY TENNISON]

Might want to Google for e2fs and inode or architecture.  I did so long ago and found that e2fs has three levels of pointers to data.  Don't know if this has anything to do with your situation but, if it is having to build more complex data structures as it goes that might account for the slow down.

Richard Geoffrion <ntlug at rain4us.net> wrote: terry wrote:
> Are we totally sure that writing the large amount of data to USB drive is
> causing the lock-up?
> Could there be some other [hardware] problem?    Maybe memory?


The PCI USB card in the server that was pulled was responsible for the 
NMIs.  Our initial test after pulling the PCI USB card resulted in a 
server lockup.  We disconnected all USB devices and cables except for 
one drive and have been testing each USB port one at a time.   So far so 
good.   We have just added in a second drive.  With only using the two 
drives we've currently tested, the system seems to be responding fast.   
A format of a 300gig USB drive is taking just over 6 minutes as opposed 
to the previous time of 3 hours.  The server hasn't locked up yet and so 
that's good.

I notice that whenever I start an e2fsck, the system seems to format the 
first 270 inodes within the first second with the rest are 
formatted/allocated/whatever at a rate that is much slower.   I'm I 
correct in thinking that the initial surge must be disk cache on the 
drive itself?

We will continue to add more USB devices after we verify the stability 
of the system after each addition.

-- 
Richard



_______________________________________________
http://www.ntlug.org/mailman/listinfo/discuss