Comments on: Lessfs-1.1.6 is available for download.

By: Fajar Priyanto

Fajar Priyanto — Fri, 17 Sep 2010 06:17:14 +0000

Ok found the problem….
The filesystem was almost full. Lessfs freeze the IO.
It runs again after I set MIN_SPACE_FREE = 5
The bad news is it rolls back to Aug 24 data …..

By: wxp

wxp — Tue, 07 Sep 2010 09:57:23 +0000

Hi，Who can tell me which functions in lessfs or others achieve the function of data dedup, thanks very much!

By: Mark

Mark — Mon, 06 Sep 2010 20:07:24 +0000

Hi Morton,

It’s not necessary to send the core file.
Just do
gdb ./lessfs /core
bt

Can you send me the output?
You can find my email address in the lessfs manual.

For people using IETD I have an experimental patch the will really speed things up when using lessfs with a 128k blocksize.

http://www.lessfs.com/wordpress/wp-content/software/iscsitarget-1.4.20.1.patch The patch is a quick hack. What it basically does is that it sends the data from the kernel to lessfs in 128k chunks instead of 4k.

Mark

By: Alex

Alex — Mon, 06 Sep 2010 12:24:59 +0000

Hi mark,

just to let you know that 1.1.6 is working great now

Well, at beginning i’ve been disappointed with some crashes using a big cache setting (2048) resulting in crashes on the test storage server (out of memory thing).

Well, it’s a dual Xeon HT 3Ghz with 6Go Ram x86 (32bit so as i have to compile TC with the good options). So settings cachesize to 256 (Mo) fixed the issue.

It is Xeon HT, i’ve got 4 cpu unit seen by the system, so i’ve set the MAX_THREADS to 20 (5x No of cpus).

Using rsync through vpn from another “normal storage server” (without lessfs), i get transfer rates between 5 to 10Mb/sec writing on lessfs storage. Which is quite good and expected i guess

So, just as an example, here is the config file :
# Automatically generated by lessfsconfig.sh
# Enable informational messages about compression.
DEBUG
# The (chopped) hashlen in bytes, minimum is 20.
HASHLEN = 24
#BLOCKDATA_IO_TYPE=file_io
BLOCKDATA_PATH=/home/replication_data/dta
BLOCKDATA_BS=13762560
#
BLOCKUSAGE_PATH=/home/replication_data/mta
BLOCKUSAGE_BS=13762560
#
DIRENT_PATH=/home/replication_data/mta
DIRENT_BS=30000000
#
FILEBLOCK_PATH=/home/replication_data/mta
FILEBLOCK_BS=132644864
#
META_PATH=/home/replication_data/mta
META_BS=30000000
#
HARDLINK_PATH=/home/replication_data/mta
HARDLINK_BS=30000000
#
SYMLINK_PATH=/home/replication_data/mta
SYMLINK_BS=30000000
#
# The freelist database is only used
# with the file_io backend
#
FREELIST_PATH=/home/replication_data/mta
FREELIST_BS=13762560
#
# CACHESIZE in MB
CACHESIZE=256
# Flush data to disk after X seconds.
COMMIT_INTERVAL=30
#
LISTEN_IP=127.0.0.1
LISTEN_PORT=100
# Not more then 2 on most machines.
MAX_THREADS=20
#DYNAMIC_DEFRAGMENTATION on or off, default is off.
DYNAMIC_DEFRAGMENTATION=on
COREDUMPSIZE=1000000000
# Consider SYNC_RELAX=1 or SYNC_RELAX=2 when exporting lessfs with NFS.
SYNC_RELAX=1
# Requires openssl and lessfs has to be configured with –with-crypto
ENCRYPT_DATA=off
# ENCRYPT_META on or off, default is off
# Requires ENCRYPT_DATA=on and is otherwise ignored.
ENCRYPT_META=on
#PASSWORD=1234
MIN_SPACE_FREE=10
#MIN_SPACE_CLEAN=15
#CLEAN_PROGRAM=/tmp/clean.sh
# Highly recommended.
ENABLE_TRANSACTIONS=on
# Select a blocksize to fit your needs.
# Most people should select 65536
BLKSIZE=65536

This is just a feedback (positive for once), in case of

System is :
Linux 2.6.32-21-generic-pae #32-Ubuntu SMP Fri Apr 16 09:39:35 UTC 2010 i686 GNU/Linux
Ubuntu 10.04.1 LTS

By: Morten

Morten — Mon, 06 Sep 2010 12:19:46 +0000

Got a couple of core file today. Have also done the following:

Changed from ietd to tgtd
installed 8GB of more memory, total is 12GB
Changed TC cache from 1GB to 4GB

Systems seems more stable, running stress test now. Will be happy to send conf and core to someone.

Morten

By: Mark

Mark — Sun, 05 Sep 2010 07:14:28 +0000

When Lessfs terminates, does it log something in /var/log/messages? Or when it does not, is there a coredump that you can backtrace? Can you email your lessfs.cfg?
How much memory do you have in this system and what is the cache size in lessfs.cfg? Can it be that the system runs out of memory?

Which iSCSI target are you using? IETD/SCST/STGT or LIO?

By: Morten

Morten — Sat, 04 Sep 2010 09:02:07 +0000

Having a new small issue with iscsi on just one of my lessfs hosts. Running esxi 4.1 with iscsi connection to dedicated hardware box, running lessfs 1.1.6.

It seems that under heavy loads I get a lot of “Abort Task” on the iscsi-target i.e :
iscsi_trgt: Abort Task (01) issued on tid:1 lun:0 by sid:4224323677913600 (Function Complete)

And after a while lessfs terminates, telling me that mount point is not connected. A restart of lessfs and iscsi-target fixes all again, but just for a while.

Wondering what can cause lessfs to terminate ? I run other hosts on the exact same configuration and the exact same load, without problems like this.

By: maru

maru — Mon, 30 Aug 2010 15:55:43 +0000

I’ll make a note to check defrag_lessfs.
There must be something wrong with it. We don’t need it anymore, but it should work anyway.

By: Morten

Morten — Mon, 30 Aug 2010 15:45:12 +0000

No core file, turned back all values to default in .config file, but still crashes. (have enabled background delete, previously)

Tried telnet localhost 100, and it is working.

Strange ……..

By: maru

maru — Mon, 30 Aug 2010 13:01:32 +0000

Not without a stack trace and a peek at your lessfs.cfg file.
I just tested it on a rather large volume myself and it works for me. ;-(

xgtest01:~ # /etc/init.d/ietd stop
Stopping iSCSI Target: Terminated
xgtest01:~ # cd /usr/src/lessfs-1.1.6
xgtest01:/usr/src/lessfs-1.1.6 # umount /fuse/
xgtest01:/usr/src/lessfs-1.1.6 # ./defrag_lessfs /etc/lessfs.cfg

Is there a core file present?
If so can you do a backtrace?
gdb ./defrag_lessfs /core
bt

P.S.

Did you know that for defragmentation you can leave lessfs mounted and just do:
telnet localhost 100
> defrag