I have been attempting to install and get OpenAFS 1.2.10 running for the past week. I spent part of the previous week going through the Gentoo and OPENAFS docs several times. It appears either the Gentoo docs are very wrong or the ebuilds are very broken. I'm willing to help fix this by running things and giving feedback but I don't have the knowledge to fix the ebuilds and make it run. 1. OpenAFS installs. 2. I am doing a server install and the docs seem to work until I have to set the acl. Then I get an error "function /afs not implemented" (see the bug on that). Looking at the Gentoo forums that error has been around for a long time. 3. If I try and start openafs with /etc/init.d/afs start I get errors. First try - nothing starts. /sbin/runscript.sh: line 40: [: /dev/loop0: binary operator expected expr: syntax error expr: syntax error * Starting AFS services... ParseCacheInfoFile: Format error in cache info file! 2 out of 3 fields successfully parsed. * Error starting AFS [ !! ] Next try: root@gandalf etc # /etc/init.d/afs start /sbin/runscript.sh: line 40: [: /dev/loop0: binary operator expected * Starting AFS services... Failed to load AFS client, not starting AFS services. /sbin/runscript.sh: line 310: [: Error Starting AFS client: integer expression expected [ !! ] /sbin/runscript.sh: line 329: return: Error: numeric argument required ParseCacheInfoFile: Format error in cache info file! 2 out of 3 fields successfully parsed. * Error starting AFS [ !! ]
Please post the result of: cat /proc/mounts cat /usr/vice/etc/cacheinfo
Here they are. The cacheinfo file I had to modify to acc the 100000 after the last : since it was blank - only the first two fields wer there. The text in /etc/afs/afs.conf says that the startup should automatically figure that but it evidently did not. OpenAFS would not start until I fixed the cacheinfo. afs:/usr/vice/cache:100000 root@gandalf etc # cat /proc/mounts rootfs / rootfs rw 0 0 /dev/root / xfs rw 0 0 none /dev devfs rw 0 0 none /proc proc rw 0 0 /dev/scsi/host0/bus0/target0/lun0/part3 /home xfs rw 0 0 /dev/scsi/host0/bus0/target0/lun0/part6 /usr xfs rw 0 0 /dev/scsi/host0/bus0/target0/lun0/part5 /files xfs rw 0 0 /dev/loop0 /usr/vice/cache ext2 rw 0 0 /dev/loop1 /vicepa ext2 rw 0 0 none /dev/shm tmpfs rw 0 0 strider:/home /mnt/strider/home nfs rw,nosuid,nodev,v2,rsize=8192,wsize=8192,hard,intr,udp,lock,addr=strider 0 0 strider:/usr/local/portage /mnt/strider/lportage nfs rw,nosuid,nodev,v2,rsize=8192,wsize=8192,hard,intr,udp,lock,addr=strider 0 0
This is almost certainly Bug 26213 again. Line 53 of /etc/init.d/afs is trying to determine the size of the partition mounted on /usr/vice/cache. The attempt fails if 'df' returns a multi-line answer, as it does when the device name is too long. What does 'df' return?
Here's df. It is multiline (I put it to a file df >tmp.txt and looked at it with a text editor. The mounts for strider are NFS mounts. root@gandalf root # df Filesystem 1K-blocks Used Available Use% Mounted on /dev/scsi/host0/bus0/target0/lun0/part2 10706684 635864 10070820 6% / /dev/scsi/host0/bus0/target0/lun0/part3 10706684 3566032 7140652 34% /home /dev/scsi/host0/bus0/target0/lun0/part6 10734620 3003984 7730636 28% /usr /dev/scsi/host0/bus0/target0/lun0/part5 220178188 25459328 194718860 12% /files /var/cache/openafs/openafs.file 298471 613 282447 1% /usr/vice/cache /files/openafs/vicepapart 4838 72 4516 2% /vicepa none 516084 0 516084 0% /dev/shm strider:/home 10706684 5414300 5292384 51% /mnt/strider/home strider:/usr/local/portage 10706652 6243904 4462748 59% /mnt/strider/lportage
I've been looking at this and it may not be the multiline bug. The code checks for the existence of /usr/vice/etc/cacheinfo. If it does NOT exist it goes through the auto creation. However, it does exist on my system. I even modifed /etc/init.d/afs to use the patch given in bug 26213 and it did not help. The fix should be included in this init file though. It looks like OpenAFS did not include it. However, if I execute PART=`cat /proc/mounts | grep vice | grep ext2 | awk '{print $1}'` $PART contains /dev/loop0 and /dev/loop1. When I start /etc/init.d/afs I get this error. /sbin/runscript.sh: line 40: [: /dev/loop0: binary operator expected Any ideas on why this is showing up? I looked at /sbin/runscript.sh but line 40 is a comment and it I can't see why any of the surrounding lines should have a problem with this.
I put a set -x in the start() function of /etc/init.d/afs and found that it's upset because $PART contains "/dev/loop0 /dev/loop1" and the -z test evidently doesn't like it (see below). I found that if I made this [ -z "$PART" ] then the test worked. This situation may not be normal but it needs to be accounted for. I was out of partitions so I created two files, make ext2 filesystems on them and then loop mounted them. With original: + PART=/dev/loop0 /dev/loop1 + '[' -z /dev/loop0 /dev/loop1 ']' /sbin/runscript.sh: line 40: [: /dev/loop0: binary operator expected Fixed: + PART=/dev/loop0 /dev/loop1 + '[' -z '/dev/loop0 /dev/loop1' ']' + :
Created attachment 23528 [details] Diff file for changes to fix $PART variable check
Now that AFS does the check and tries to start I get to this point. What next? root@gandalf init.d # /etc/init.d/afs start * Starting AFS services... afsd: All AFS daemons started. afsd: Can't mount AFS on /afs(22) * Error starting AFS [ !! ]
That problem could be a lot of things, probably not related to the ebuild. Make sure the kernel sources you compiled against match your running kernel. Make sure the directory /afs exists. Check /usr/vice/etc/ThisCell and /usr/vice/etc/CellServDB.
Is this bug still valid?
AFS not mounting is probably due to following the gentoo AFS documentation. It is unclear on a number of things. Unless you have created your root.afs and root.cell entries properly /afs cannot mount. /afs is a "magic" mount point. To work around this try the following: Assuming you have created your "admin" id: klog admin vos create root.afs <servername> <vice partition name> vos create root.cell <servername> <vice partition name> now restart your AFS process (you may have to reboot) and continue with the Gentoo documentation.
Okay. I may try that. I have had to drop working on getting OpenAFS running for a while due to other things taking my time. I got OpenAFS to start/run - kind of - by dumping the Gentoo docs and going to the OpenAFS docs. I got everything created and ready to run. I also had to fix the /etc/init.d/afs file because a fix for multiline fstab output wasn't in there (it had gotten lost). I may just unmerge OpenAFS and then start over again. Thanks.
OpenAFS needs a developer to take up maintenance.
Per comment #8 and #11: make sure bosserver is running when you attempt to start afsd. PATH=/usr/afs/bin:$PATH bosserver -noauth & bos adduser aquarius mmokrejs.admin -cell doma # the next commands make bosserver immediately start server processes bos create aquarius ptserver simple /usr/afs/bin/ptserver -cell doma bos create aquarius vlserver simple /usr/afs/bin/vlserver -cell doma bos create aquarius fs fs /usr/afs/bin/fileserver /usr/afs/bin/volserver /usr/afs/bin/salvager -cell doma # bos stop -instance fs -server aquarius # bos delete -instance fs -server aquarius pts createuser mmokrejs -cell doma pts createuser mmokrejs.admin -cell doma pts adduser mmokrejs.admin system:administrators -cell doma pts membership mmokrejs.admin -cell doma vos create aquarius /vicepa root.afs -cell doma mkdir -p /afs chmod a+rx /afs mkdir -p /usr/vice/cache /usr/vice/etc/afsd -nosettime -verbose & If this works, reboot the machine because: bosserver runs in insecure mode and you cannot easily get rid of the afsd kernel process. after the machine comes up, start bosserver without the "-noauth" option, start again afsd and continue: for kth-krb4 do: kauth mmokrejs.admin for heimdal do: kauth mmokrejs/admin ls -la /afs df vos listvol aquarius vos listvldb fs sa /afs system:anyuser rl fs sa /afs system:authuser rl fs examine /afs vos create aquarius /vicepa root.cell -cell doma fs mkmount /afs/doma root.cell fs setacl /afs/doma system:anyuser rl fs setacl /afs/doma system:authuser rl fs mkmount /afs/.doma root.cell -rw fs setacl /afs/.doma system:anyuser rl fs setacl /afs/.doma system:authuser rl fs examine /afs vos addsite aquarius /vicepa root.afs vos addsite aquarius /vicepa root.cell vos release root.afs vos release root.cell fs checkvolumes ls -la /afs vos release root.cell fs checkvolumes vos release root.afs
New ebuilds for openafs 1.2.13 (stable) and 1.3.85 (experimental) are available for testing. According to openafs-ml, 1.3.85 is currently undergoing testing so it can become 1.4rc. The new init script doesn't make a /etc/openafs/cacheinfo by itself anymore, the administrator is supposed to create it. Also the check if the cache-location contains a mounted ext2-partition has been removed (cache doesn't need to be a seperate partition, much less can it only be ext2, care with 1.2.x is probably advisable though) It'd be great if the new proposed ebuilds could be tested to determine the current degree of brokenness of "OpenAFS on Gentoo" as mentioned in the bug's summary :)
Problems were most likely caused by misconfiguration by the user.
