Re: [Kosmosfs-users] KFS errors
Status: Alpha
Brought to you by:
sriramsrao
Menu
▾
▴
Re: [Kosmosfs-users] KFS errors
|
From: Sriram R. <sri...@gm...> - 2008-09-20 20:19:59
|
The config files look good. For the two computers over the wireless setup, can you ping from one to another and see how long that takes? Well...here is one possibility: - you write to the file - there is a valid write lease on that file - when you open the file for reading, the metaserver can't hand out a read lease (since there is a valid write lease) - the lease expires; which takes 60 secs - the client asks for the read lease and gets it. Alternately, you could try writing a bigger file (something on the order of a 1G or so) and try reading it. Sriram On Sat, Sep 20, 2008 at 1:01 PM, Mateusz Berezecki <mat...@gm...> wrote: > On Sat, Sep 20, 2008 at 8:59 PM, Mateusz Berezecki <mat...@gm...> wrote: >> On Sat, Sep 20, 2008 at 8:48 PM, Sriram Rao <sri...@gm...> wrote: >>> Hi Mateusz, >> >> Hi Sriram, > > To make things funnier, here are new stats > > m:tools m$ ./kfsping -m -s monkey.local -p 20000 > Up servers: 1 > s=monkey , p=30000, total=3(GB), used=1.39698e-08(GB), > util=4.65661e-07%, nblocks=2, lastheard=5 (sec), ncorrupt=0 > > m:tools m$ ./kfsping -c -s monkey.local -p 20000 > Meta-server: 0 > Total-space: 0 (MB) > Used-space: 0 (MB) > > The config files are correct this time I believe. Writing works > locally, but very few operations > work over the network. I am connected to the wireless lan with my 2 computers. > I have uploaded one file from the same server the master and chunk > servers are running, > and try to download it from remote computer. Here's what I got > > 09-20-2008 21:55:50.327 DEBUG - > (/Users/m/Downloads/kfs-0.2.1/src/cc/libkfsClient/KfsClient.cc:2351) > file-id for dir: /foo/ (file = bar.txt) is 18 > 09-20-2008 21:55:50.333 DEBUG - > (/Users/m/Downloads/kfs-0.2.1/src/cc/libkfsClient/KfsClient.cc:1881) > Fileid: 17, # of chunks: 0 > Connect: : Operation timed out > 09-20-2008 21:57:05.009 DEBUG - > (/Users/m/Downloads/kfs-0.2.1/src/cc/libkfsClient/KfsClient.cc:1902) > Size of kfsfid = 17, size = 0 > Time to open KFS file : 75 seconds > > The computers are next to each other... > > > Any ideas? > > Mateusz >>> Can you get me the output of KfsPing and your config files? Each >>> chunkserver node tells the metaserver how space it is exporting and >>> that is likely too small. >> >> $ ./kfsping -m -s localhost -p 20000 >> Up servers: 1 >> s=monkey , p=30000, total=3(MB), used=0(MB), util=0%, nblocks=0, >> lastheard=1 (sec), ncorrupt=0 >> >> $ ./kfsping -c -s localhost -p 20000 >> Meta-server: 0 >> Total-space: 0 (MB) >> Used-space: 0 (MB) >> >> ChunkServer.prp >> >> 1 chunkServer.metaServer.hostname = 192.168.0.146 >> 2 chunkServer.metaServer.port = 20100 >> 3 chunkServer.clientPort = 30000 >> 4 chunkServer.clusterKey = test-cluster >> 5 chunkServer.rackId = 0 >> 6 chunkServer.totalSpace = 314572800 >> 7 chunkServer.chunkDir = /home/mateusz/kosmos/fs1/bin/kfschunk >> 8 chunkServer.logDir = /home/mateusz/kosmos/fs1/bin/kfslog >> >> >> MetaServer.prp >> 1 metaServer.clientPort = 20000 >> 2 metaServer.chunkServerPort = 20100 >> 3 metaServer.clusterKey = test-cluster >> 4 metaServer.cpDir = /home/mateusz/kosmos/meta/bin/kfscp >> 5 metaServer.logDir = /home/mateusz/kosmos/meta/bin/kfslog >> >> machines.cfg >> 1 [metaserver] >> 2 node: 192.168.0.146 >> 3 rundir: /home/mateusz/kosmos/meta >> 4 baseport: 20000 >> 5 clusterkey: test-cluster >> 6 >> 7 [chunkserver1] >> 8 node: 192.168.0.146 >> 9 rundir: /home/mateusz/kosmos/fs1 >> 10 baseport: 30000 >> 11 space: 300 M >> >> >> I think these are all quite normal settings? >> >> Mateusz >> >>> >>> What happens on writes: >>> - each chunk is 64MB in size >>> - the metaserver hands out a write lease to one of the chunkservers >>> and space is reserved as long as the write lease is out >>> - if your system has less than 250MB per chunkserver, writes wil >>> likely fail (in the code, we check that if a node is over 90% full, we >>> don't send writes to it); >>> >>> Sriram >>> >>> On Sat, Sep 20, 2008 at 10:57 AM, Mateusz Berezecki <mat...@gm...> wrote: >>>> Hello list readers, >>>> >>>> I have compiled the latest KFS 0.2.1 and here's my experience so far >>>> (with the fresh clean KFS setup): >>>> >>>> KFS Path: /bar.txt >>>> 09-20-2008 19:45:09.498 DEBUG - >>>> (/Users/m/Downloads/kfs-0.2.1/src/cc/libkfsClient/KfsClient.cc:2351) >>>> file-id for dir: / (file = bar.txt) is 2 >>>> 09-20-2008 19:45:09.503 DEBUG - >>>> (/Users/m/Downloads/kfs-0.2.1/src/cc/libkfsClient/KfsClient.cc:1792) >>>> getalloc: fid=11 offset: 0 failed with code(-2): No such file or >>>> directory >>>> 09-20-2008 19:45:09.504 DEBUG - >>>> (/Users/m/Downloads/kfs-0.2.1/src/cc/libkfsClient/KfsClient.cc:1505) >>>> LocateChunk (-2): No such file or directory >>>> Assertion failed: (bytes_read > 0), function DownloadFile, file >>>> /Users/m/work/rss/cluster/KfsBroker.cc, line 18. >>>> Abort trap >>>> >>>> The first line, the assertion come from my program which uses libkfsClient. >>>> I have created a file on the filesystem and wanted to read it programatically. >>>> >>>> The problem is that reading did not work as expected. >>>> So I figured out I would try going with tests under bin/tests directory. >>>> >>>> The PerfWriter test gives me the following output. >>>> mateusz@monkey:~/src/kfs-0.2.1/bin/tests$ ./KfsPerfWriter -p kfs.props >>>> -m 3 -b 2 -f /test >>>> Doing writes to: /test # MB = 3 # of bytes per write: 2 >>>> 09-20-2008 19:45:48.498 DEBUG - >>>> (/home/mateusz/src/kfs-0.2.1/src/cc/libkfsClient/KfsClient.cc:409) >>>> Connecting to metaserver at: localhost:20000 >>>> Making dir: >>>> Mkdir returned: 0 >>>> 09-20-2008 19:45:48.526 DEBUG - >>>> (/home/mateusz/src/kfs-0.2.1/src/cc/libkfsClient/KfsClient.cc:2351) >>>> file-id for dir: / (file = test) is 2 >>>> 09-20-2008 19:45:48.527 DEBUG - >>>> (/home/mateusz/src/kfs-0.2.1/src/cc/libkfsClient/KfsClient.cc:1792) >>>> allocate: fid=19 offset: 0 failed with code(-28): No space left on >>>> device >>>> 09-20-2008 19:45:48.527 DEBUG - >>>> (/home/mateusz/src/kfs-0.2.1/src/cc/libkfsClient/KfsWrite.cc:141) >>>> ----Write done: asked: 2, got: 0----- >>>> Write rate: 0 (Mbps) >>>> Write rate: 0 (MBps) >>>> >>>> >>>> Below is my df -h output >>>> >>>> Filesystem Size Used Avail Use% Mounted on >>>> /dev/sda1 359G 50G 291G 15% / >>>> varrun 2.0G 112K 2.0G 1% /var/run >>>> varlock 2.0G 0 2.0G 0% /var/lock >>>> udev 2.0G 52K 2.0G 1% /dev >>>> devshm 2.0G 24K 2.0G 1% /dev/shm >>>> lrm 2.0G 44M 1.9G 3% >>>> >>>> So what is the no space left on device message all about? >>>> >>>> ------------------------------------------------------------------------- >>>> This SF.Net email is sponsored by the Moblin Your Move Developer's challenge >>>> Build the coolest Linux based applications with Moblin SDK & win great prizes >>>> Grand prize is a trip for two to an Open Source event anywhere in the world >>>> http://moblin-contest.org/redirect.php?banner_id=100&url=/ >>>> _______________________________________________ >>>> Kosmosfs-users mailing list >>>> Kos...@li... >>>> https://lists.sourceforge.net/lists/listinfo/kosmosfs-users >>>> >>> >> > |