Page 1 of 1
Fatdog64-v5.11 filesystem limited to 2 GB? (Solved)
Posted: Sun 30 Jan 2011, 21:10
by gcmartin
I run a LiveDVD FATDOG on a system with 8GB of RAM (soon to go to 24GB). Just noticed today that it appears the FATDOG filesystem is limited to 2GB.
Code: Select all
# df -m
Filesystem 1M-blocks Used Available Use% Mounted on
tmpfs 1932 451 1481 24% /initrd/pup_rw
tmpfs 1932 445 1487 23% /initrd/pup_ro1
tmpfs 184 183 2 100% /initrd/mnt/tmpfs
/dev/loop0 182 182 0 100% /initrd/pup_ro2
unionfs 1932 451 1481 24% /
Question: Is this true? Is this an artificial limitation or is it because of something else? Is there some item I'm suppose to do at boot-time so that FATDOG uses all available RAM? (seems odd)
Is this documented somewhere?
Edited: Here's a 2nd system....same issue
Code: Select all
-Computer-
Processor : 2x AMD Athlon(tm) 64 X2 Dual Core Processor 4600+
Memory : 4054MB (524MB used)
Operating System : Puppy Linux 0.50
User Name : root (root)
Date/Time : Sun 30 Jan 2011 04:16:56 PM GMT+5
Thanks in advance
Posted: Sun 30 Jan 2011, 23:17
by jamesbond
Hmmm that's odd. Please tell me the output of running these commands from terminal:
Code: Select all
free
mount -a tmpfs tmpfs /mnt/data
df -h
umount /mnt/data
Posted: Mon 31 Jan 2011, 04:00
by kirk
Are you running a multi-session live DVD or just booted ram only? It looks like the init script tries to use 1/2 of ram for tmpfs in those cases.
Reply for more info -Fatdog64-v5.11 filesystem limited to 2G
Posted: Mon 31 Jan 2011, 17:47
by gcmartin
@JamesBond
Code: Select all
# free
total used free shared buffers
Mem: 4054748 3181696 873052 0 28728
Swap: 0 0 0
Total: 4054748 3181696 873052
# mount -a tmpfs tmpfs /mnt/data
Yields a usage error
# df -h
Filesystem Size Used Avail Use% Mounted on
tmpfs 1.9G 1.1G 886M 55% /initrd/pup_rw
tmpfs 1.9G 445M 1.5G 23% /initrd/pup_ro1
tmpfs 184M 183M 1.6M 100% /initrd/mnt/tmpfs
/dev/loop0 182M 182M 0 100% /initrd/pup_ro2
unionfs 1.9G 1.1G 886M 55% /
# umount /mnt/data
not mounted
# mount
rootfs on / type rootfs (rw,relatime)
tmpfs on /initrd/pup_rw type tmpfs (rw,relatime,size=1977376k)
tmpfs on /initrd/pup_ro1 type tmpfs (rw,relatime,size=1977376k)
tmpfs on /initrd/mnt/tmpfs type tmpfs (rw,relatime,size=188064k)
/dev/loop0 on /initrd/pup_ro2 type squashfs (ro,noatime)
unionfs on / type aufs (rw,relatime,si=99a6850ede9486f6)
none on /proc type proc (rw,relatime)
none on /dev/pts type devpts (rw,relatime,gid=2,mode=620)
none on /sys type sysfs (rw,relatime)
none on /proc/bus/usb type usbfs (rw,relatime)
@Kirk, yes I am running a multi-session LiveDVD.
Update
Posted: Mon 31 Jan 2011, 20:28
by gcmartin
I noticed that my swap device was NOT in use. I restarted the LiveCd system with the swap partition visable. Things have changed.
Code: Select all
# mount
rootfs on / type rootfs (rw,relatime)
tmpfs on /initrd/pup_rw type tmpfs (rw,relatime,size=4143940k)
tmpfs on /initrd/mnt/tmpfs type tmpfs (rw,relatime,size=188064k)
/dev/loop0 on /initrd/pup_ro2 type squashfs (ro,noatime)
unionfs on / type aufs (rw,relatime,si=ab32d1b41aa556bc)
none on /proc type proc (rw,relatime)
none on /dev/pts type devpts (rw,relatime,gid=2,mode=620)
none on /sys type sysfs (rw,relatime)
none on /proc/bus/usb type usbfs (rw,relatime)
# free
total used free shared buffers
Mem: 4054748 442848 3611900 0 46044
Swap: 8450112 0 8450112
Total: 12504860 442848 12062012
# df -h
Filesystem Size Used Avail Use% Mounted on
tmpfs 4.0G 12M 4.0G 1% /initrd/pup_rw
tmpfs 184M 183M 1.6M 100% /initrd/mnt/tmpfs
/dev/loop0 182M 182M 0 100% /initrd/pup_ro2
unionfs 4.0G 12M 4.0G 1% /
I think that explains this issue of why memory was being used the way it was. We now see 4GB in use on the FS.
2nd question: My swap partition is 4GB (4138MB from Gparted). Does FATDOG compress the Swap FS? (
or is this some sort of a reporting error)
P.S. PDiag script was helpful in getting to the information which provided clues. Useful tool. Maybe in the next upgrade.
Posted: Mon 31 Jan 2011, 22:31
by jamesbond
Typo ... the "mount -a" should "mount -t" ... everything else is the same. But don't bother to run that command again. I don't have a machine with 8GB worth of RAM, but this doesn't look right. Can you tell me the output of
?
cheers!
Posted: Mon 31 Jan 2011, 23:29
by gcmartin
Here is my 4GB system.
Code: Select all
-Computer-
Processor : 2x AMD Athlon(tm) 64 X2 Dual Core Processor 4600+
Memory : 4054MB (524MB used)
Operating System : Puppy Linux 0.50
User Name : root (root)
# cat /proc/meminfo | grep Total
MemTotal: 4054748 kB
SwapTotal: 8450112 kB
VmallocTotal: 34359738367 kB
# free
total used free shared buffers
Mem: 4054748 1178152 2876596 0 27496
Swap: 8450112 0 8450112
Total: 12504860 1178152 11326708
# mount -t tmpfs tmpfs /mnt/data
# df -h
Filesystem Size Used Avail Use% Mounted on
tmpfs 4.0G 163M 3.8G 5% /initrd/pup_rw
tmpfs 1.9G 445M 1.5G 23% /initrd/pup_ro1
tmpfs 184M 183M 1.6M 100% /initrd/mnt/tmpfs
/dev/loop0 182M 182M 0 100% /initrd/pup_ro2
unionfs 4.0G 163M 3.8G 5% /
tmpfs 2.0G 0 2.0G 0% /mnt/data
My 1.5GB 32bit Intel laptop running Wary without a SWAP seems to shop same symptoms as originally reported here. This behavior may be built in for some reason. But, when they see a swap, things change.
Found swap report issue
Posted: Tue 01 Feb 2011, 00:53
by gcmartin
OK, I found the source of the swap report I showed in the prior pst.
I actually had plugged a USB into the system before shutdown last night. I did not unplug it when I restarted the system with a HDD configured with SWAP.
So I had 2 swap devices available to the system (each 4GB) when I rebooted and began data capture. Sorry for the report on double swap size. Found the problem when I didn't see the same doubling on a 32bit Wary laptop.
But, the original finding of using ONLY 1/2 of total memory happens exactly on Wary as it does on FATDOG. If you do NOT have a swap partition, Puppies will NOT make all of your memory available to the filesystem.
This fact will be useful as I begin to change 2011 layouts of the Puppies on my LAN that I use and test with.
FATDOG is still the only Puppy that I have which is a competitor with Vista/System7. Its the only one where SAMBA 3.5.6 runs with little effort in setup and running. Other Puppies were cripple by 5.2 development soon after Mobeus got it working in 5.1 . We have gotten lots of hits by person interested in FATDOG+SAMBA.
I had suggested at the beginning of the year that I expected to begin work on using FATDOG as a PDC to avoid a $3000+ cost vs M$ (server and client licensing in a 25 users LAN). I am starting that today and hope to have a simple implementation documented for FATDOG before end of the month. (To me, the documentation is
always the most important.)
Again, I thank all who did raise an eyelid when I opened this post. I can only guess why memory is dissected, but, understanding what causes the issue is very important to me.
Code: Select all
# free
total used free shared buffers
Mem: 4054748 1002056 3052692 0 27120
Swap: 4233120 0 4233120
Total: 8287868 1002056 7285812
# cat /proc/meminfo | grep Total
MemTotal: 4054748 kB
SwapTotal: 4233120 kB
VmallocTotal: 34359738367 kB
# df -m
Filesystem 1M-blocks Used Available Use% Mounted on
tmpfs 3998 41 3958 2% /initrd/pup_rw
tmpfs 1932 445 1487 23% /initrd/pup_ro1
tmpfs 184 183 2 100% /initrd/mnt/tmpfs
/dev/loop0 182 182 0 100% /initrd/pup_ro2
unionfs 3998 41 3958 2% /
Posted: Tue 01 Feb 2011, 09:20
by jamesbond
gcmartin, I'm a bit confused. In the first post you mention that your system has 8GB of RAM. In the following posts you mention that your system only have 4GB of RAM. Can you confirm which one? RAM = memory only. Don't count the extra memory you get from swap - swap is swap and it isn't RAM.
"cat /proc/meminfo | grep Total" ==> this command shows that the kernel's understanding of system memory. In all your report it shows 4GB (less a little, which is ok). If your machine really have 8GB of RAM (excluding swap), this is really disconcerting because that means the kernel isn't seeing the rest of your RAM.
Now the /pup_rw filesystem - as kirk said, the size of this is approximately half of what you see in the firstline, first column output of the "free" command. If your machine has only as 4GB RAM, I am now amazed that this /pup_rw is 4GB - it should only 2 GB, regardless how how many swap you have.
After discovery, I Only used the 4GB system for diagnostics
Posted: Tue 01 Feb 2011, 20:38
by gcmartin
The diagnostics we did on this and the commands which I used to gather information was done using the 4GB system. The 8GB system was being tested to understand how FATDOG saw its SCSIs when I discovered the memory issue. That system, a Microsoft server, has been put back into production. And was only used briefly.
Again, all of the data gather and shown here is from the 4GB system.
jamesbond wrote:gcmartin, I'm a bit confused. In the first post you mention that your system has 8GB of RAM. In the following posts you mention that your system only have 4GB of RAM.
...
Now the /pup_rw filesystem - as kirk said, the size of this is approximately half of what you see in the firstline, first column output of the "free" command. If your machine has only as 4GB RAM, I am now amazed that this /pup_rw is 4GB - it should only 2 GB, regardless how how many swap you have.
As I see this, it appears that ALL PUPs, thus far, seem to show 1/2 of memory allocated to filesystem iff no SWAP partition is available. Finding a SWAP partition, PUPs seemingly expand memory use to all available RAM.
An external view of this seems that this is a reasonable approach. The 64bit system with SWAP does seem snappier as I load, test, and run applications (but this is emotional as I don't have tools to validate this or track this under loads - nor do I have testbank for FATDOG/PUPs to do some sort of consistent measurements).
Again this behavior has been observed in both FATDOG and WARY on differing platforms (i.e. desktops and laptops).
Hope this explains my observations, thus far.