NSLU2-Linux
view · edit · print · history

Debian.DebianInstallerDisconnectsSSH History

Hide minor edits - Show changes to markup

August 17, 2008, at 10:27 AM by tk -- reply
Changed lines 22-23 from:
to:

-- reply J.S: only had about 9.8mb free ram on trying this - still crashes. linux pc formatting using mke2fs isnt detected by the slug :(

February 26, 2008, at 04:41 PM by Del Merritt -- Not always a memory problem!
Changed lines 3-4 from:

This is usually a symptom of the NSLU2 running out of available memory. When a process grows too much, the OOM (Out Of Memory) killer of the kernel kills it, and its parents. The OOM killer is controlled with /proc/sys/vm/overcommit_memory and /proc/sys/vm/overcommit_ratio. A given process can be made immune to the OOM killer with echo OOM_DISABLE > /proc/<pid>/oom_adj. Although it is interesting to know about the OOM killer, there is no report that tweaking it solves the problem.

to:

The simplest cause of this is the configuration of the ssh client used to connect to the NSLU2 (slug) in question. See Debian bug 458154 for details and discussion. In summary, you may be able to get away with adding the following to your ~/.ssh/config:

ServerAliveInterval 3

Otherwise this may be a symptom of the NSLU2 running out of available memory. When a process grows too much, the OOM (Out Of Memory) killer of the kernel kills it, and its parents. The OOM killer is controlled with /proc/sys/vm/overcommit_memory and /proc/sys/vm/overcommit_ratio. A given process can be made immune to the OOM killer with echo OOM_DISABLE > /proc/<pid>/oom_adj. Although it is interesting to know about the OOM killer, there is no report that tweaking it solves the problem.

January 13, 2008, at 03:51 PM by Loic Dachary --
Changed lines 84-89 from:

~ # dmesg | tail -5 707 slab pages 4690 pages shared 371 pages swap cached Out of Memory: Kill process 2176 (sshd) score 72 and children. Out of memory: Killed process 2229 (sshd).

to:

~ # dmesg Adding 503608k swap on /dev/sda2. Priority:-2 extents:1 across:503608k oom-killer: gfp_mask=0x200d0, order=0 Mem-info: DMA per-cpu: cpu 0 hot: high 6, batch 1 used:0 cpu 0 cold: high 2, batch 1 used:1 DMA32? per-cpu: empty Normal per-cpu: empty HighMem? per-cpu: empty Free pages: 1076kB (0kB HighMem?) Active:1876 inactive:4012 dirty:3 writeback:3138 unstable:0 free:269 slab:725 mapped:63 pagetables:69 DMA free:1076kB min:724kB low:904kB high:1084kB active:7504kB inactive:16048kB present:32768kB pages_scanned:1376 all_unreclaimable? no lowmem_reserve[]: 0 0 0 0 DMA32? free:0kB min:0kB low:0kB high:0kB active:0kB inactive:0kB present:0kB pages_scanned:0 all_unreclaimable? no lowmem_reserve[]: 0 0 0 0 Normal free:0kB min:0kB low:0kB high:0kB active:0kB inactive:0kB present:0kB pages_scanned:0 all_unreclaimable? no lowmem_reserve[]: 0 0 0 0 HighMem? free:0kB min:128kB low:128kB high:128kB active:0kB inactive:0kB present:0kB pages_scanned:0 all_unreclaimable? no lowmem_reserve[]: 0 0 0 0 DMA: 25*4kB 2*8kB 2*16kB 1*32kB 4*64kB 1*128kB 0*256kB 1*512kB 0*1024kB 0*2048kB 0*4096kB = 1076kB DMA32?: empty Normal: empty HighMem?: empty Swap cache: add 2193, delete 2171, find 546/561, race 0+0 Free swap = 499188kB Total swap = 503608kB Free swap: 499188kB 8192 pages of RAM 376 free pages 712 reserved pages 725 slab pages 3192 pages shared 22 pages swap cached Out of Memory: Kill process 2176 (sshd) score 74 and children. Out of memory: Killed process 6189 (sshd).

January 13, 2008, at 03:35 PM by Loic Dachary -- instructions can fail
Changed lines 12-13 from:
  1. Tell it to use the swap space as swap, it will need to format the swap space as swap, and there is not way to prevent this formatting (that I can tell). Tell it to mount the root partition as / and in the same screen select Do not format / keep existing data. Do not let it do any extra work. it will mount these, unmount swap, remount swap, and let you proceed through the installation.
to:
  1. Tell it to use the swap space as swap, it will need to format the swap space as swap, and there is not way to prevent this formatting (that I can tell).
  2. Tell it to mount the root partition as / and in the same screen select Do not format / keep existing data. Do not let it do any extra work. it will mount these, unmount swap, remount swap, and let you proceed through the installation.
Added lines 23-90:

Note by Loic Dachary: the instructions suggesting to add swap prior to running mke2fs on a large partition succeeded on one NSLU2 and always fails on another one. I'm stuck with it:

~ # cat /proc/swaps
Filename                                Type            Size    Used    Priority
~ # mkswap /dev/sda1
Setting up swapspace version 1, size = 515682304 bytes
~ # swapon /dev/sda1
~ # mke2fs -j /dev/sda2
mke2fs 1.40-WIP (14-Nov-2006)
Filesystem label=
OS type: Linux
Block size=4096 (log=2)
Fragment size=4096 (log=2)
921120 inodes, 1839516 blocks
91975 blocks (5.00%) reserved for the super user
First data block=0
Maximum filesystem blocks=1887436800
57 block groups
32768 blocks per group, 32768 fragments per group
16160 inodes per group
Superblock backups stored on blocks:
        32768, 98304, 163840, 229376, 294912, 819200, 884736, 1605632

Writing inode tables: Connection to 192.168.1.130 closed by remote host.
Connection to 192.168.1.130 closed.
(reconnect)
~ # cat /proc/meminfo
MemTotal:        29988 kB
MemFree:          9676 kB
Buffers:             0 kB
Cached:          10956 kB
SwapCached:       2528 kB
Active:           9020 kB
Inactive:         6380 kB
HighTotal:           0 kB
HighFree:            0 kB
LowTotal:        29988 kB
LowFree:          9676 kB
SwapTotal:      503592 kB
SwapFree:       501064 kB
Dirty:               0 kB
Writeback:           0 kB
AnonPages:        2096 kB
Mapped:           2300 kB
Slab:             2448 kB
PageTables:        276 kB
NFS_Unstable:        0 kB
Bounce:              0 kB
CommitLimit:    518584 kB
Committed_AS:     8392 kB
VmallocTotal:   991232 kB
VmallocUsed:     17084 kB
VmallocChunk:   966652 kB
~ # cat /proc/swaps
Filename                                Type            Size    Used    Priority
/dev/scsi/host0/bus0/target0/lun0/part1 partition       503592  2528    -1
~ # cat /proc/sys/vm/overcommit_memory
0
~ # cat /proc/sys/vm/overcommit_ratio
50
~ # dmesg | tail -5
707 slab pages
4690 pages shared
371 pages swap cached
Out of Memory: Kill process 2176 (sshd) score 72 and children.
Out of memory: Killed process 2229 (sshd).
January 13, 2008, at 11:14 AM by Loic Dachary -- the problem exists with RC2 also
Changed lines 1-6 from:

The current installer for NSLU2 available via http://www.slug-firmware.net (debian-etch-rc1-20061102.zip) seems to repeatedly close SSH during installation, causing the install to restart. This is usually a symptom of the NSLU2 running out of available memory.

When a process grows too much, the OOM (Out Of Memory) killer of the kernel kills it, and its parents. The OOM killer is controlled with /proc/sys/vm/overcommit_memory and /proc/sys/vm/overcommit_ratio. A given process can be made immune to the OOM killer with echo OOM_DISABLE > /proc/<pid>/oom_adj. Although it is interesting to know about the OOM killer, there is no report that tweaking it solves the problem.

  • I verified that this problem still exists on RC2 firmware -- Paul Tan.
to:

The current installer for NSLU2 available via http://www.slug-firmware.net (debian-etch-rc1-20061102.zip and debian-4.0r2.zip) seems to repeatedly close SSH during installation, causing the install to restart.

This is usually a symptom of the NSLU2 running out of available memory. When a process grows too much, the OOM (Out Of Memory) killer of the kernel kills it, and its parents. The OOM killer is controlled with /proc/sys/vm/overcommit_memory and /proc/sys/vm/overcommit_ratio. A given process can be made immune to the OOM killer with echo OOM_DISABLE > /proc/<pid>/oom_adj. Although it is interesting to know about the OOM killer, there is no report that tweaking it solves the problem.

January 13, 2008, at 11:10 AM by Loic Dachary --
Changed lines 3-4 from:

When a process grows too much, the OOM (Out Of Memory) killer of the kernel kills it, and its parents. The OOM killer is controlled with /proc/sys/vm/overcommit_memory and /proc/sys/vm/overcommit_ratio. A given process can be made immune to the OOM kill with echo OOM_DISABLE > /proc/<pid>/oom_adj. Although it is interesting to know about the OOM killer, there is no report that tweaking it solves the problem.

to:

When a process grows too much, the OOM (Out Of Memory) killer of the kernel kills it, and its parents. The OOM killer is controlled with /proc/sys/vm/overcommit_memory and /proc/sys/vm/overcommit_ratio. A given process can be made immune to the OOM killer with echo OOM_DISABLE > /proc/<pid>/oom_adj. Although it is interesting to know about the OOM killer, there is no report that tweaking it solves the problem.

January 13, 2008, at 11:09 AM by Loic Dachary -- OOM
Added lines 3-4:

When a process grows too much, the OOM (Out Of Memory) killer of the kernel kills it, and its parents. The OOM killer is controlled with /proc/sys/vm/overcommit_memory and /proc/sys/vm/overcommit_ratio. A given process can be made immune to the OOM kill with echo OOM_DISABLE > /proc/<pid>/oom_adj. Although it is interesting to know about the OOM killer, there is no report that tweaking it solves the problem.

May 19, 2007, at 03:08 PM by Christopher Salinardi -- More info on partitioning and formatting
Changed lines 16-17 from:

Note by Chris S: I got it to work on a 500GB partition by using "modprobe -r" to unload the fat and vfat drivers (lsmod to view the modules loaded). They're a waste of memory if you aren't installing to a FAT partition. Then I hit <Go Back> on the partitioner to have it close the partitioner and return to the installer menu (this also saves memory). All told, I had 11MB of free ram when I entered the mke2fs command.

to:

Note by Chris S: I got it to work on a 500GB partition by using "modprobe -r" to unload the fat and vfat drivers (lsmod to view the modules loaded). They're a waste of memory if you aren't installing to a FAT partition. Also use "modprobe -r" to unload the ixp4xx_beeper driver to save a little more RAM. Then hit <Go Back> on the partitioner to have it close the partitioner and return to the installer menu (this also saves memory). All told, I had over 11MB of free ram when I entered the mke2fs command. FYI, when creating the partitions DO NOT USE sector 0. Fdisk will let you use it, but when you reboot your drive will be useless until you are able to recover the boot sector.

May 14, 2007, at 01:35 AM by Chris Salinardi -- Added instructions to reduce memory usage and format large partitions
Added lines 16-17:

Note by Chris S: I got it to work on a 500GB partition by using "modprobe -r" to unload the fat and vfat drivers (lsmod to view the modules loaded). They're a waste of memory if you aren't installing to a FAT partition. Then I hit <Go Back> on the partitioner to have it close the partitioner and return to the installer menu (this also saves memory). All told, I had 11MB of free ram when I entered the mke2fs command.

April 30, 2007, at 08:34 PM by Milam --
Added lines 16-17:

So it seems that it is the formatting of the root / primary partition that causes the memory crash, rather than the formatting of the swap space.

April 30, 2007, at 08:33 PM by Milam --
Changed lines 12-13 from:
  1. tell it to use the swap space as swap. tell it to mount the root partition as /. do not let it do extra work. it will mount these, unmount swap, remount swap, and let you proceed through the installation.
to:
  1. Tell it to use the swap space as swap, it will need to format the swap space as swap, and there is not way to prevent this formatting (that I can tell). Tell it to mount the root partition as / and in the same screen select Do not format / keep existing data. Do not let it do any extra work. it will mount these, unmount swap, remount swap, and let you proceed through the installation.
Changed line 17 from:
  1. Use your desktop Linux computer to partition and format the disk prior to installation
to:
  1. Use your desktop Linux computer to partition and format the disk prior to installation
April 05, 2007, at 08:05 PM by fcarolo -- link cleanup
Changed lines 1-4 from:

The current installer for NSLU2? available via http://www.slug-firmware.net (debian-etch-rc1-20061102.zip) seems to repeatedly close SSH during installation, causing the install to restart. This is usually a symptom of the NSLU2 running out of available memory.

  • I verified that this problem still exists on RC2? firmware -- Paul Tan.
to:

The current installer for NSLU2 available via http://www.slug-firmware.net (debian-etch-rc1-20061102.zip) seems to repeatedly close SSH during installation, causing the install to restart. This is usually a symptom of the NSLU2 running out of available memory.

  • I verified that this problem still exists on RC2 firmware -- Paul Tan.
April 04, 2007, at 05:40 PM by jabdebianorg --
Changed lines 14-17 from:

Resizing the SSH terminal screen will also cause a disconnect; don't do that.

Attempting to format very large partitions also creates this problem, but the workaround does not help. Try to make the primary partition smaller. This will allow the installer to run in the normal fashion.

to:

Resizing the SSH terminal screen will also cause a disconnect; don't do that. Attempting to format very large partitions also creates this problem, but the workaround does not help. Try to make the primary partition smaller. This will allow the installer to run in the normal fashion.

April 04, 2007, at 05:40 PM by jabdebianorg --
Changed lines 16-19 from:

Attempting to format very large partitions also creates this problem, but the workaround does not help. Try to make the primary partition smaller. This will allow the installer to run in the normal fashion.

to:

Attempting to format very large partitions also creates this problem, but the workaround does not help. Try to make the primary partition smaller. This will allow the installer to run in the normal fashion.

An alternate solution is to:

  1. Use your desktop Linux computer to partition and format the disk prior to installation
March 28, 2007, at 03:44 AM by ptan --
Added lines 3-4:
  • I verified that this problem still exists on RC2? firmware -- Paul Tan.
March 17, 2007, at 05:13 PM by Nick Solonar --
Added lines 13-14:

Attempting to format very large partitions also creates this problem, but the workaround does not help. Try to make the primary partition smaller. This will allow the installer to run in the normal fashion.

January 12, 2007, at 02:25 PM by Dean Jackson --
Added lines 1-12:

The current installer for NSLU2? available via http://www.slug-firmware.net (debian-etch-rc1-20061102.zip) seems to repeatedly close SSH during installation, causing the install to restart. This is usually a symptom of the NSLU2 running out of available memory.

A solution to this is to:

  1. run the installer up to the point where it will create partitions for you.
  2. log into the box via a second terminal, and choose the second option, a command shell.
  3. use fdisk to partition the disk you intend to use, including at least 128 MB of swap space.
  4. use mkswap /dev/<swap partition> and then swapon /dev/<swap partition> to activate that swap space.
  5. use mke2fs -j /dev/<root partition> to format your root partition.
  6. logout, and go back to the SSH session still connected to the installer.
  7. tell it to use the swap space as swap. tell it to mount the root partition as /. do not let it do extra work. it will mount these, unmount swap, remount swap, and let you proceed through the installation.

Resizing the SSH terminal screen will also cause a disconnect; don't do that.

view · edit · print · history · Last edited by tk.
Based on work by Del Merritt, Loic Dachary, Christopher Salinardi, Chris Salinardi, Milam, fcarolo, jabdebianorg, ptan, and Nick Solonar.
Originally by Dean Jackson.
Page last modified on August 17, 2008, at 10:27 AM