Author |
Message |
red321
|
Posted: Thu May 18, 2006 11:33 am |
|
Joined: Sun Mar 07, 2004 5:34 am
Posts: 116
Location:
UK
|
|
Top |
|
|
Greg Frost
|
Posted: Sat May 20, 2006 5:12 pm |
|
Joined: Mon May 10, 2004 8:08 pm
Posts: 1891
Location:
Adelaide, Australia
|
I ran out of inodes on my /myth partition when I attempted to extract the tarball of the frontend into /myth/terminal_server/nfsroot.
How did you get around this?
From my backend:
Code: root@mythtv:/myth/terminal_server# df -i Filesystem Inodes IUsed IFree IUse% Mounted on /dev/hda1 500992 149001 351991 30% / /dev/hda3 47168 2221 44947 5% /myth
From my frontend: Code: root@fedell:~# df -i Filesystem Inodes IUsed IFree IUse% Mounted on /dev/hda1 500992 116063 384929 24% / /dev/hda3 1024 587 437 58% /myth 10.1.1.6:/myth/video 47168 2221 44947 5% /myth/video
By deleting everything in /usr/src on the frontend, I got the inode requirement down from 116063 to 74000, but that is still not going to fit on my /myth partition.
Did you originally do a manual install on your backend?
/usr alone uses around 57000 inodes, so if you could find a way to mount the /usr directory on the backend instead of having a full copy of it, that would help.
|
|
Top |
|
|
red321
|
Posted: Sun May 21, 2006 2:28 am |
|
Joined: Sun Mar 07, 2004 5:34 am
Posts: 116
Location:
UK
|
Hmm,
Sorry Greg, not a problem I ran into, as I have upgraded so many times that my machine isn't a "clean install". I have moved /myth/tv off onto another physical disk, and my /myth partition has plenty of inodes to spare. Therefore didnt even see this one coming
Code: /dev/hda6 16400384 600750 15799634 4% /myth
The easy solution is to reformat your /myth partition with lots more inodes
A cleverer option would be to image your .usr directory, and then mount it as a loop. Apparently /usr is supposed to be read-only anyway. Kind of spoils the idea of not imaging things though.
You could also symlink /usr to another partition, or move the whole root. Nothing magic about where you put it, as long as it is exported.
[I havent proved this, so you might need to experiment]
So no reason for you not symlink to the backends /usr directory, as long as you export it. However be aware you will then be able to "mess" with the backend from the frontend.
post back if you get it working.
|
|
Top |
|
|
Greg Frost
|
Posted: Sun May 21, 2006 2:38 am |
|
Joined: Mon May 10, 2004 8:08 pm
Posts: 1891
Location:
Adelaide, Australia
|
Quote: So no reason for you not symlink to the backends /usr directory, as long as you export it. However be aware you will then be able to "mess" with the backend from the frontend. Thanks, I had already started attempting to do this. Quote: edit /etc/fstab
remove any refrences to local disks, as on the frontend there arnt going to be any ! remove
/dev/hda1 / ext3 defaults,errors=remount-ro 0 1
replace with:
192.168.254.4:/myth/terminal_server/nfsroot / nfs defaults,auto,noatime 0 2
I assume that this bit is really asking me to edit /myth/terminal_server/nfsroot/etc/fstab and not /etc/fstab.
If that is the case, it might be worth clarifying it on the wiki before someone hoses their backend.
|
|
Top |
|
|
red321
|
Posted: Sun May 21, 2006 2:42 am |
|
Joined: Sun Mar 07, 2004 5:34 am
Posts: 116
Location:
UK
|
Good point !
I'll fix the Wiki
I tried symlinking /myth/terminal_services/nfsroot/usr to /usr. and exporting /usr. It didnt work for me because my backend installs and frontends are sufficiently diffrent now for ther eto be symlinks between /usr and other parts of the backend filesystem, which dont apear on the frontend filesystem. Bit of a managment nightmare to intertwine 2 file systems
|
|
Top |
|
|
Greg Frost
|
Posted: Sun May 21, 2006 3:02 am |
|
Joined: Mon May 10, 2004 8:08 pm
Posts: 1891
Location:
Adelaide, Australia
|
My tftp on the frontend timed out. When I went back to check /tftpboot/pxelinux.cfg/default on the backend, I found it wasn't there! On closer inspection it appears that /tftpboot was a symlink to /tmp/tftpboot, so rebooting the backend blew away the configuration. I ran prep-ts again and this time removed the tftpboot symlink and moved /tmp/tftpboot to /tftpboot.
I'll reconfigure and let you know how I get on.
I'm a little confused about this: Quote: ALLOW Root access - If you dont do this, you will end up with a Read Only access to the root filesystem, and it will all go horribly wrong. I didnt see any option to "allow root access" when running prep-ts.
|
|
Top |
|
|
Greg Frost
|
Posted: Sun May 21, 2006 3:17 am |
|
Joined: Mon May 10, 2004 8:08 pm
Posts: 1891
Location:
Adelaide, Australia
|
Well, I still get a TFTP open timeout. You say tftp should be taken care of by inetd. Looking at /etc/inetd.conf, there is a line for tftpd that is commented out. Quote: #tftp dgram udp wait nobody /usr/sbin/tcpd /usr/sbin/in.tftpd /boot Do I need to uncomment this and configure tftp somehow?
This is what I see in /var/log/syslog: Code: May 21 18:52:08 mythtv dhcpd: DHCPDISCOVER from 00:06:5b:7c:8e:5e via eth0 May 21 18:52:09 mythtv dhcpd: DHCPOFFER on 10.1.1.250 to 00:06:5b:7c:8e:5e via eth0 May 21 18:52:10 mythtv dhcpd: DHCPREQUEST for 10.1.1.250 (10.1.1.6) from 00:06:5b:7c:8e:5e via eth0 May 21 18:52:10 mythtv dhcpd: DHCPACK on 10.1.1.250 to 00:06:5b:7c:8e:5e via eth0 May 21 18:52:10 mythtv in.tftpd[5272]: connect from 10.1.1.250 (10.1.1.250) May 21 18:52:11 mythtv in.tftpd[5273]: cannot set groups for user nobody May 21 18:52:12 mythtv in.tftpd[5274]: cannot set groups for user nobody May 21 18:52:16 mythtv in.tftpd[5275]: cannot set groups for user nobody May 21 18:52:22 mythtv in.tftpd[5276]: cannot set groups for user nobody May 21 18:52:30 mythtv in.tftpd[5277]: cannot set groups for user nobody May 21 18:52:40 mythtv in.tftpd[5278]: cannot set groups for user nobody May 21 18:53:16 mythtv in.tftpd[5279]: cannot set groups for user nobody May 21 18:54:16 mythtv ntpd[4838]: synchronized to LOCAL(0), stratum 13 May 21 18:54:16 mythtv ntpd[4838]: kernel time sync disabled 0041 May 21 18:54:28 mythtv in.tftpd[5280]: cannot set groups for user nobody May 21 18:55:20 mythtv ntpd[4838]: synchronized to 218.214.129.61, stratum 2
|
|
Top |
|
|
Greg Frost
|
Posted: Sun May 21, 2006 3:49 am |
|
Joined: Mon May 10, 2004 8:08 pm
Posts: 1891
Location:
Adelaide, Australia
|
Ok, I got past that problem. I needed to edit /etc/inetd.conf and replace Code: #tftp dgram udp wait nobody /usr/sbin/tcpd /usr/sbin/in.tftpd /boot with Code: tftp dgram udp wait root /usr/sbin/tcpd /usr/sbin/in.tftpd -s /tftpboot Then restart inetd. I have updated the wiki accordingly.
I am getting a kernel panic now because it can't mount the root filesystem, but I'll keep looking to see if I have hammed something up.
|
|
Top |
|
|
red321
|
Posted: Sun May 21, 2006 8:28 am |
|
Joined: Sun Mar 07, 2004 5:34 am
Posts: 116
Location:
UK
|
Shows why good How-tos are hard to find
I ran this on my own main system, then ran a clean install on another machine, but only as a autoupdate. This leaves /myth on the root partition if you dont allready have one.
Thats why I didnt spot the inodes gotcha.
Not sure why you are seeing a kernel panic.
If you have got that far , then I guess:
DHCP, PXE, TFTP are all working. You should be seeing a NFS access request in your syslog, unless:
1) You dont have nfs-kernel-server running
2) You are not exporting the root filesystem.
3) your default config is wrong.
Cant think of anything else.
Oh yes Ican. Did you read the "dont run knoppix-terminalserver" ever again, ever, I really mean it bit
You might be running with th eoriginal vmlinuz file, rather than the new shiny one that you built with NFSroot support. The big problem her eis that as soon as you run Terminalservices it happily undoes all your good work. In fact I just di that myself while checking out one of your questions
Quote: 'm a little confused about this: Quote: ALLOW Root access - If you dont do this, you will end up with a Read Only access to the root filesystem, and it will all go horribly wrong. I didnt see any option to "allow root access" when running prep-ts.
the actual wording is [] secure Disable root access for clients.
That adds the word SECURE to the append options in the default pxelinux.cfg. It makes the NFSroot ro effectively. However as this is now irrelevant, as we rwrite a complete new append line, I will remove it from the wiki.
[ I was originally trying to be less intrusive, and us the treminalservices setup to configure some of the options I have now done by hand. Gave up in the end ]
|
|
Top |
|
|
mgrassi99
|
Posted: Thu Jun 15, 2006 12:47 am |
|
Joined: Wed Jan 28, 2004 12:04 am
Posts: 68
|
GREAT howto; I had already been playing around with TFTP and LTSP and other diskless solutions without success, and this got me over the hump.
One hiccup I encountered using the NFSroot method: my kernel loaded, and the setup scripts started to execute, but I got loads of:
Code: nfs: RPC call returned error 101 RPC: sendmsg call returned error 101
errors, and the system would hang. After a few google searches, I came upon an article that said when using an NFS root you must NOT re-run your /etc/init.d/networking script as you've already brought your net interface up by that point, and by running that you actually take your interface down briefly thus losing your root FS. I got around this by simply adding a "exit;" at the top of the NFS root's /etc/init.d/networking script, and all is well. Not sure if anyone else has seen this...
-Mike
|
|
Top |
|
|
Greg Frost
|
Posted: Sat Jun 17, 2006 1:03 am |
|
Joined: Mon May 10, 2004 8:08 pm
Posts: 1891
Location:
Adelaide, Australia
|
I had a go at getting this going again. Because of my inode problem on the /myth partition, I just put the NFS root on the root partition of the backend. That worked fine. I then ran into the problem that mgrassi99 had. Thanks for posting a solution. That worked a treat for me.
Now I'm trying to work out how to get the frontend to mount the backend's /usr directory so that I dont need a whole extra copy of it (its 1.2G which takes a fair chunk out of the root partition).
|
|
Top |
|
|
Greg Frost
|
Posted: Sat Jun 17, 2006 1:57 am |
|
Joined: Mon May 10, 2004 8:08 pm
Posts: 1891
Location:
Adelaide, Australia
|
So here's what I did:
In the nfs root: rename usr to usr2 so it will not be used. Create an empty usr directory to be used as a mount point.
In /etc/exports on the backend, add this: Code: /usr *(ro) In etc/fstab in the nfsroot add this line: Code: 10.1.1.6:/usr /usr nfs defaults,auto,noatime 0 2 Now when I netboot the frontend, I see some errors during the initial stages of the boot where /usr is not available. Then, after X starts and when the frontend would normally show up, I get the xterm asking me to enter the root password that you get when you do an install. I suspect that some of the tuner detection code that checks if a card has been added or removed is not working properly which is causing that to run.
The good news is, if I quit that xterm and start the frontend, it just works. So I guess I'm almost there if I can resolve the issue of mounting /usr soon enough.
Does anyone know what during the boot process causes the additional nfs mount points to be mounted?
|
|
Top |
|
|
red321
|
Posted: Thu Jun 29, 2006 3:30 pm |
|
Joined: Sun Mar 07, 2004 5:34 am
Posts: 116
Location:
UK
|
Greg,
Quick fix would be to edit ~/.fluxbox/apps
change out the knoppmyth-run entry (cant rememner the exact syntax)
change it to just mythfrontend
Then your login willl not be reqested, as that is due to the knoppmyth-run script detecting ( or not detecting) flags that it expects set.
|
|
Top |
|
|
Greg Frost
|
Posted: Thu Jun 29, 2006 9:40 pm |
|
Joined: Mon May 10, 2004 8:08 pm
Posts: 1891
Location:
Adelaide, Australia
|
Thanks, but I just changed the startup numbers for mountall and mountnfs in /etc/rcS.d so that they were before the knoppmyth script that checks for tuners and it all worked beautifully.
|
|
Top |
|
|
sinspot1
|
Posted: Sun Jul 09, 2006 2:27 pm |
|
Joined: Thu Dec 22, 2005 5:32 pm
Posts: 198
|
I had a question...... I am setting up the disklessfe and was wondering to all of the FE have to use the same image or could I have different ones. For example that one FE is set to use the digital out and another is using the analog outs or one is using the S-VIDEO and the other is using the DVI out. You see what I mean.
Thanks
_________________ Marty
Long live Myth and Knoppmyth!!!!!
|
|
Top |
|
|