chunk server failed to start

[龙斌]

dear all,

I have configured the kfs-0.5 metaserver on a machines and a sigle chunk server on another machine.

when I use the script python kfslaunch.py -f machines.cfg  -s try to start the chunkserver and metaserver, I find it that the mataserver succeed to start ,but the chunkserver failed to start.

In the kfs's  log file of chunk server,I can see a filed with postfix .pid which contained a integer number.I guess this may be the pid of the chunk server service process, but I find that there isn't a process of this pid.

Can you help me~thank you

[Tony Qin]

You may forget to present machine ip list as input parameter.

As for running on single/multiple node, you can use:

python kfslaunch.py -f machines.cfg -m machines.txt --start

[stackpop]

I have configured the chunkserver in machine.cfg ,not in a txt file.
Even if i changed the machine.cfg and create a chunkserver list file
like you, the problem still exist.

[Tony Qin]

If your configuration is correct. maybe you should check your hostname.

[Tony Qin]

In my single/multiple-node testbed, it works fine.

As for webui, you should copy the machines.txt to webui directory
before you run kfssetup. Or, the webui can not find the servers.

Good Luck,

Tony

On Mon, May 23, 2011 at 10:05 PM, stackpop <t...@stackpop.org> wrote:

[Tony Qin]

If still fail, please open and paste the log.

2011/5/23 Tony Qin <anqi...@gmail.com>:

[Tony Qin]

Could you open and paste the log?

On Mon, May 23, 2011 at 10:05 PM, stackpop <t...@stackpop.org> wrote:

[stackpop]

in the logs directory,there is only a chunkserver.pid file.where can I
find more logs information?

[Tony Qin]

vim kfsrun.sh
...
bin/$server $config $SERVER_LOG_FILE # > /dev/null 2>&1 &
...

disable the "> /dev/null 2>&1 &"

run again the kfslaunch.py, you will get chunkserver output.

since the chunkserver will lookup the metaserver hostname, please
confirm your hostname is configured correctly.

2011/5/24 stackpop <t...@stackpop.org>:

[shining]

The contents of the SERVER_LOG_FILE is:

[192.168.0.1]: Starting metaserver...
[192.168.0.1]: Starting cleaner...
[192.168.0.2]: Starting chunkserver...
[192.168.0.2]: Loading key chunkServer.metaServer.hostname with value
192.168.0.1
[192.168.0.2]: Loading key chunkServer.metaServer.port with value
20100
[192.168.0.2]: Loading key chunkServer.clientPort with value 30000
[192.168.0.2]: Loading key chunkServer.clusterKey with value meta-
cluster
[192.168.0.2]: Loading key chunkServer.rackId with value 0
[192.168.0.2]: Loading key chunkServer.md5sum with value
8f6dae8d09275951dbc7cb88eec6fb68
[192.168.0.2]: Loading key chunkServer.totalSpace with value
10737418240
[192.168.0.2]: Loading key chunkServer.chunkDir with value /home/kfs/
chunk/bin/kfschunk
[192.168.0.2]: Loading key chunkServer.logDir with value /home/kfs/
chunk/bin/kfslog
[192.168.0.2]: Loading key chunkServer.loglevel with value INFO
[192.168.0.2]: Using chunk server client port: 30000
[192.168.0.2]: Using chunk dir = /home/kfs/chunk/bin/kfschunk
[192.168.0.2]: Using log dir = /home/kfs/chunk/bin/kfslog
[192.168.0.2]: Total space = 10737418240
[192.168.0.2]: cleanup on start = 0
[192.168.0.2]: Chunk server rack: 0
[192.168.0.2]: using cluster key = meta-cluster

It seems nothing wrong~

[Tony Qin]

r u in single node?

what are the relationship of 192.168.0.1 and 192.168.0.2?

2011/5/24 shining <shini...@gmail.com>:

[shining]

192.168.0.1 metaserver
192.168.0.2 chunkserver

I install kfs in two separated machines.

On 5月24日, 下午1时01分, Tony Qin <anqin....@gmail.com> wrote:
> r u in single node?
>
> what are the relationship of 192.168.0.1 and 192.168.0.2?
>

> 2011/5/24 shining <shining....@gmail.com>:

[shining]

My machines.cfg file:

# KFS Machine configuration file
[metaserver]
node: 192.168.0.1
rundir: /home/kfs/meta
baseport: 20000
backup_path: 192.168.0.1:/home/kfs/meta_bkup
clusterkey: meta-cluster
loglevel: INFO
numservers: 2

[chunkserver1]
node: 192.168.0.2
rundir: /home/kfs/chunk
chunkDir: /home/kfs/chunk/bin/kfschunk
baseport: 30000
space: 10 G
loglevel: INFO

[Tony Qin]

In 192.168.0.2, I guess you miss the installation of libboost_regex.
Both machines have to be install the necessary required libs.

You can make sure this by:

vim kfsrun.sh
...
bin/$server $config $SERVER_LOG_FILE # > /dev/null 2>&1 &
...

disable the "> /dev/null 2>&1 &" and run the kfsrun.sh to what is output.

2011/5/24 shining <shini...@gmail.com>:

[shining]

my os is centos 5.5

I've already done "yum install boost boost-devel"
Isn't it enough?Do I need extra package of libboost_regex which is
not included in boost?

On 5月24日, 下午1时49分, Tony Qin <anqin....@gmail.com> wrote:
> In 192.168.0.2, I guess you miss the installation of libboost_regex.
> Both machines have to be install the necessary required libs.
>
> You can make sure this by:
>
> vim kfsrun.sh
> ...
> bin/$server $config $SERVER_LOG_FILE # > /dev/null 2>&1 &
> ...
>
> disable the "> /dev/null 2>&1 &" and run the kfsrun.sh to what is output.
>

> 2011/5/24 shining <shining....@gmail.com>:

[jrckkyy]

http://blog.csdn.net/jrckkyy/archive/2010/11/24/6032877.aspx

http://blog.csdn.net/jrckkyy/archive/2011/01/09/6125786.aspx

2011/5/23 龙斌 <t...@stackpop.org>

--

 

此致

       敬礼！

[Tony Qin]

good summary.

But it seems that miss the deployment on cluster, :-)

2011/5/24 jrckkyy <jrc...@gmail.com>:

[stackpop]

大家都是中国人，打一段中文吧。
楼上提供的安装方法，和我之前参考的安装文档没有差别。
我这边metaserver已经启动成功了，可以看到20000端口的进程metaserver，还有http的web界面也可以访问了。
chunkserver那边的logs文件夹每次执行启动命令都会新建好一个pid的文件，但是却发现当前进程中没有一个这样的pid的文件。

On 5月24日, 下午2时40分, Tony Qin <anqin....@gmail.com> wrote:
> good summary.
>
> But it seems that miss the deployment on cluster, :-)
>

> 2011/5/24 jrckkyy <jrck...@gmail.com>:

[binlong]

The problem still exist...help are eargly in need....

On 5月24日, 下午2时40分, Tony Qin <anqin....@gmail.com> wrote:

> good summary.
>
> But it seems that miss the deployment on cluster, :-)
>

> 2011/5/24 jrckkyy <jrck...@gmail.com>:

>
>
>
> >http://blog.csdn.net/jrckkyy/archive/2010/11/24/6032877.aspx
> >http://blog.csdn.net/jrckkyy/archive/2011/01/09/6125786.aspx
> > 2011/5/23 龙斌 <t...@stackpop.org>
>
> >> dear all,
>
> >> I have configured the kfs-0.5 metaserver on a machines and a sigle chunk
> >> server on another machine.
>
> >> when I use the script python kfslaunch.py -f machines.cfg -s try to start
> >> the chunkserver and metaserver, I find it that the mataserver succeed to
> >> start ,but the chunkserver failed to start.
>
> >> In the kfs's log file of chunk server,I can see a filed with postfix .pid
> >> which contained a integer number.I guess this may be the pid of the chunk
> >> server service process, but I find that there isn't a process of this pid.
>
> >> Can you help me~thank you
>
> > --
>
> > 此致

> > 敬礼！- 隐藏被引用文字 -
>
> - 显示引用的文字 -