/usr became full while updating via cvs

[IronForge]

Greetings:

Hope you all had a great holiday break.

I was getting an update of -current using cvs. I used one server for the src tree, which rejected me when I tried to update the ports tree.

I then closed the terminal thread, opened another window, and checked out a tree from scratch, and updated src and ports.

Rant into some problems when I tried to update xenocara:

/usr: write failed, file system is full
cvs [checkout aborted]: writing xenocara/app/rgb/config.guess: No space left on device
...looked around to see how much space I had..

# df -hi
Filesystem   Size     Used     Avail Cap'ty      iused   ifree  %iused  Mounted on
/dev/wd0a 1005M 55.1M   899M       6%      2972 152930    2%   /
/dev/wd0k  18.1G  279M   16.9G       2%      2447 2414063  0%   /home
/dev/wd0d    2.6G  140K     2.5G        0%          32  363742    0%   /tmp
/dev/wd0f     2.0G   2.0G  -92.0M  105%  277972      7850  97%   /usr
/dev/wd0g1005M  170M    785M    18%    9132    146770    6%   /usr/X11R6
/dev/wd0h   3.5G   2.1G    1.2G        63%   77745  415949  16%   /usr/local
/dev/wd0j    1.9G   973M    892M    52%   33619   226219  13%   /usr/obj
/dev/wd0i    1.9G   753M    1.1G      40%   87448   172390  34%   /usr/src
/dev/wd0e   4.1G   118M    3.8G        3%   12945   558701    2%   /var

I used disklabel -A for my wd0 drive. I've another HDD (100GB); but I've never gotten around to adding it.
Q1: Are there any "caches" to clear out so I can continue?

I'm not a partition expert, so I'd like some r-e-m-e-d-i-a-l level help on the following:
Q2a: If I need to reconfigure (to expand /usr), what's the best way to do it w/o having to recalc - will the new disklabel (will RTFM) allow a "hard" allocation for a drive section during an "auto" reconfiguration?
Q2b: If I were to add the 2nd HDD, may I have a few examples of what a "good" partition allocation of a 40/100GB looks like? (e.g., can I "split" /home btwn the 2 drives, or "move" it to occupy the second HDD? etc.)

Thank you, all.

[ocicat]

Originally Posted by IronForge

Are there any "caches" to clear out so I can continue?

Judicious use of du(1) will help you answer your own question.

Given the continuing issues you have experienced with Xenocara (& because the df(1) output above does not show mountpoints for /usr/xenocara & /usr/xobj...), both /usr/xenocara & /usr/xobj will reside in your /usr partition. Given that the df(1) output also shows this partition to only be 2GB in size, this is most likely the cause for the /usr partition to fill up knowing what little I do about your past experiences.

In other words, the source to Xenocara & the artifacts of building may be consuming /usr.

At one time, I use to create individual partitions for /usr/xenocara & /usr/xobj which were 2GB each. However, this chews up two partitions, & OpenBSD currently has the limitation of 16 partitions total per drive (including partition'c'...). Now I simply create a larger /usr partition.

You may find that creating partitions for /usr/xenocara & /usr/xobj as being a viable solution, however, no one other than you knows everything you may have in /usr. 2GB for /usr seems rather small. Creating additional partitions for /usr/xenocara & /usr/xobj might solve your problem, but it might also only mask /usr being so meager.

Again, I can only guess given my limited knowledge of what you are doing.

If I need to reconfigure (to expand /usr), what's the best way to do it...

Study growfs(8). However, growfs(8) requires that free space be found adjacent to the existing partition. Not knowing how you partitioned your disk (study the output of disklabel(8)...), I suspect that /usr is sandwiched between other partitions. If this is the case:

• you might be able to shuffle partitions about backing one up followed by restoring it elsewhere, but you may or may not have another drive available to serve as temporary storage, or it may just be impossible to do so altogether. Shuffling is also involved enough that you may make a catastrophic mistake midway.
• count this as a learning experience & reinstall.

If I were to add the 2nd HDD, may I have a few examples of what a "good" partition allocation of a 40/100GB looks like?

A good partitioning scheme is one that works, what works is highly dependent upon usage patterns. Since no two people use OpenBSD in the same way, it is very difficult to suggest or dictate how you should partition your drive. Because of the vagaries anyone can experience (like that described above...), jggimi frequently makes the suggestion that new users simply allocate one large partition for / until users have had enough experience to understand their own specific needs.

...can I "split" /home btwn the 2 drives, or "move" it to occupy the second HDD?

Yes, /home can be split or moved.

1. Study Section 14.3.
2. Partition the disk as desired using:
   • fdisk(8) if necessary.
   • disklabel(8).
   • newfs(8).
3. mount(8) the disk.
4. cp(1) or mv(1) files as needed Depending upon permission issues, dump(8) & restore(8) may be better choices.
5. Make appropriate changes to /etc/fstab.

Again, consider the chances for error in such a manual process. Irregardless of what choice you make, back up everything of importance first, & consider reinstallation as being a prudent option.

[IronForge]

Ocicat:

Will get to it.

[jggimi]

Originally Posted by ocicat

Because of the vagaries anyone can experience (like that described above...), jggimi frequently makes the suggestion that new users simply allocate one large partition for / until users have had enough experience to understand their own specific needs.

Yes, I do. Using the automatic configuration values is often fine, but not always. Admins who are building -current components are supposed to have either sufficient capacity management knowledge or disk space, to avoid storage capacity problems. As an expedient, the savvy admin can always make temporary filesystems for use during building, using MFS, loopback NFS mounts, external USB drives, etc.

(It is not strictly required to build Xenocara to stay -current on a workstation. One can, instead, use the x*tgz filesets from the most recent snapshot, unless one has special needs, or just wants to gain experience.)



I just installed -current on a new netbook over the holidays, and selected the automatic disklabel configuration, just because I'd never used one before. In the event I need to build something, and have insufficient disk space in a partitition, I'll use one of those expediencies. So far, anything I've had to build for it (a couple of ports without packages due to licensing) I've built on another platform.

I did make -two- partition changes from the automatic default, after installation. /home was converted to a softraid(4) crypto discipline device, and /tmp was replaced with an MFS partition. If I ever use an application that saves possibly private information into /var/tmp, I will make it an encrypted partition as well.