Issues moving large number of files

gauge73 · 05-07-2008, 07:57 AM

Hey all,

I'm going to give you the short version of this long-winded problem I've been having. Initially, I was trying to back up a bunch of files from a SATA drive to an IDE drive on the same Suse 10 machine. During the copy (using bacula, rsync, or just cp) the machine would hang. It took about 10 minutes to log in with SSH and unmount the drive. I pretty much had to reboot at that point.

I switched the IDE drive to a USB bay. Same issue. I've since bought a SATA drive so I'm copying from SATA to SATA. Same issue. I've tried ReiserFS and ext3. Same issue.

So, I've found that if I boot the machine in "fail safe" mode in GRUB, the backup runs fine with rsync, but it's quite slow. I'm watching my MP3s back up now, and it's maybe 1 file every second or two. So, that's roughly 3MB/s or so. That's quite slow for a local SATA to SATA copy.

I'm not really sure what all the options in the kernel line of the "Fail Safe" entry in GRUB mean. Any ideas which of these resolved the issue? Any ideas if one of those options would be causing the slow copy speed, or do you think it's just the nature of rsync (which I've never used)?

Thanks very much in advance for any responses. This issue is killing me.

Randux · 05-07-2008, 09:23 AM

rsync is as fast as cp, it's just a wrapper. There's no reason rsync should be slow unless it's during the phase that it's figuring out what to copy. Note that if you have fancy control statements for rsync and you are changing files on the drive at the same time, rsync will try to make sure it does what you want and it *can* slow down.

do rsync with the -v option (or maybe even -vv) and you can see where the delay is.

gauge73 · 05-07-2008, 09:45 AM

The options I'm using on the rsync command are -vxrpogtc.

I started looking through the kernel parms used in the fail safe grub entry, and I noticed two that seem relevent; edd=off and ide=nodma. I'm sure the ide=nodma is going to have a very serious impact on speed. The only question is whether or not that option is the same one that kept it from locking up on me.

I just rebooted the box with edd=off but without ide=nodma. I'm re-running the backup and will see if it runs any faster and if it crashes again.

Randux · 05-07-2008, 02:47 PM

Having DMA off will ruin your performance for sure as you said. I don't know what all those rsync options are without looking at the man page again. I use -ax which I believe includes *most* of your -xrpogtc but not all of them.

gauge73 · 05-07-2008, 10:45 PM

Well, this issue continues to piss in my Cherios, it seems. I've eliminated rsync as a possible issue as the problem occurs with a simple cp -Rv now. Here are some errors showing up in the messages file that are meaningless to me but might shed some light on the subject for you gurus:

Code:

May  7 22:35:14 fileserver kernel: sdb: Write Protect is off
May  7 22:35:14 fileserver kernel: sdb: Mode Sense: 00 3a 00 00
May  7 22:35:14 fileserver kernel: SCSI device sdb: drive cache: write back
May  7 22:37:25 fileserver kernel: ata1.01: limiting speed to PIO2
May  7 22:37:25 fileserver kernel: ata1.01: exception Emask 0x0 SAct 0x0 SErr 0x0 action 0x2 frozen
May  7 22:37:25 fileserver kernel: ata1.01: cmd 24/00:00:3f:64:d5/00:04:00:00:00/f0 tag 0 cdb 0x0 data 524288 in
May  7 22:37:25 fileserver kernel:          res 50/01:01:01:00:00/01:00:00:00:00/10 Emask 0x202 (HSM violation)
May  7 22:37:25 fileserver kernel: ata1: soft resetting port
May  7 22:37:26 fileserver kernel: ata1.00: configured for UDMA/133
May  7 22:37:26 fileserver kernel: ata1.01: configured for PIO2
May  7 22:37:26 fileserver kernel: ata1: EH complete
May  7 22:37:26 fileserver kernel: SCSI device sda: 586114704 512-byte hdwr sectors (300091 MB)
May  7 22:37:26 fileserver kernel: sda: Write Protect is off
May  7 22:37:26 fileserver kernel: sda: Mode Sense: 00 3a 00 00
May  7 22:37:26 fileserver kernel: SCSI device sda: drive cache: write back
May  7 22:37:26 fileserver kernel: SCSI device sdb: 586114704 512-byte hdwr sectors (300091 MB)
May  7 22:37:26 fileserver kernel: sdb: Write Protect is off
May  7 22:37:26 fileserver kernel: sdb: Mode Sense: 00 3a 00 00
May  7 22:37:26 fileserver kernel: SCSI device sdb: drive cache: write back
May  7 22:37:26 fileserver kernel: SCSI device sda: 586114704 512-byte hdwr sectors (300091 MB)
May  7 22:37:26 fileserver kernel: sda: Write Protect is off
May  7 22:37:26 fileserver kernel: sda: Mode Sense: 00 3a 00 00
May  7 22:37:26 fileserver kernel: SCSI device sda: drive cache: write back
May  7 22:37:26 fileserver kernel: SCSI device sdb: 586114704 512-byte hdwr sectors (300091 MB)
May  7 22:37:26 fileserver kernel: sdb: Write Protect is off
May  7 22:37:26 fileserver kernel: sdb: Mode Sense: 00 3a 00 00
May  7 22:37:26 fileserver kernel: SCSI device sdb: drive cache: write back
May  7 22:40:20 fileserver gconfd (root-5045): starting (version 2.12.1), pid 5045 user 'root'
May  7 22:40:21 fileserver gconfd (root-5045): Resolved address "xml:readonly:/etc/opt/gnome/gconf/gconf.xml.mandatory" to a read-only configuration source at position 0
May  7 22:40:21 fileserver gconfd (root-5045): Resolved address "xml:readwrite:/root/.gconf" to a writable configuration source at position 1
May  7 22:40:21 fileserver gconfd (root-5045): Resolved address "xml:readonly:/etc/opt/gnome/gconf/gconf.xml.defaults" to a read-only configuration source at position 2
May  7 22:42:07 fileserver smbd[3968]: [2008/05/07 22:42:07, 0] printing/print_cups.c:cups_cache_reload(85)
May  7 22:42:07 fileserver smbd[3968]:   Unable to connect to CUPS server localhost - Connection refused
May  7 22:42:07 fileserver smbd[3968]: [2008/05/07 22:42:07, 0] printing/print_cups.c:cups_cache_reload(85)
May  7 22:42:07 fileserver smbd[3968]:   Unable to connect to CUPS server localhost - Connection refused

Tinkster · 05-08-2008, 04:57 AM

IDE-/SATA chipset, kernel version?

gauge73 · 05-08-2008, 08:15 AM

This is Suse 10.1 with kernel version 2.6.16.46-0.12-bigsmp. The mobo is an ABIT IC7-MAX3 with the 875P chipset which uses the ICH5R south bridge which includes an SiS SATA controller. I hope this is enough info. Thanks for the replies!

Tinkster · 05-08-2008, 01:23 PM

Can you post the output of lspci?

And I'm not too familiar with SuSEs kernel version numbers,
but that looks dreadfully old to me (going by the stock
kernel 2.6.25.2 I'm using), the last (official) 2.6.16 was
published in Apr 2006. And I seem to recall that around
that version number some bug was introduced that not only
caused slow transfers under certain circumstance, but also
data corruption on some chipsets (my boss was telling me
that he had the problem).

Cheers,
Tink