[OpenAFS] OpenAFS Project List

Nathan Neulinger nneul@umr.edu
Tue, 13 Mar 2001 19:24:51 -0600


> |------------------+--------------+---------------+----------------------|
> | Groups in Groups | In Progress  |               | Peter Honeyman       |
> |                  |              |               | (UMich).             |
> |------------------+--------------+---------------+----------------------|

Would also like to see a cell-wide cleanacl/cleangroups to clear out
old/deleted principals/subgroups/etc.


> |------------------+--------------+---------------+----------------------|
> | Foreign Principal| In Progress ?| N/A           | Derrick Brashear     |
> | in a pts group   | Available    |               | (CMU) will retrieve  |
> |                  |              |               | this information from|
> |                  |              |               | the info-afs archive.|
> |                  |              |               | Derrick will also    |
> |                  |              |               | publish information  |
> |                  |              |               | on how to set up a   |
> |                  |              |               | cross-cell           |
> |                  |              |               | environment in       |
> |                  |              |               | general.             |
> |------------------+--------------+---------------+----------------------|

Isn't that all automatic with aklog? About the only thing that I could
see here would be giving the 'fs' command the ability to auto-create the
principals the same way that aklog does.

> |------------------+--------------+---------------+----------------------|
> | AFSDB Resource   |              | High          |                      |
> | Record           |              |               |                      |
> | Identification   |              |               |                      |
> | (using DNS)      |              |               |                      |
> |------------------+--------------+---------------+----------------------|

AFSDB or SRV/TXT or both?

> |------------------+--------------+---------------+----------------------|
> | Large            |              | High, but     | Peter Honeyman       |
> | files/volumes (> |              | difficult     | (UMich).  Note:      |
> | 2 GB files, > 8  |              |               | Peter will get work  |
> | GB volumes)      |              |               | related to this      |
> |                  |              |               | effort from Antonelli|
> |                  |              |               | & his group at UMich |
> |                  |              |               | .  Hartmut Reuter    |
> |                  |              |               | (Max Planck) may also|
> |                  |              |               | have some related    |
> |                  |              |               | work.                |
> |------------------+--------------+---------------+----------------------|

Out of curiousity, why are 8+ GB volumes a problem currently? (I'm
thinking I have one in my cell is why, although it might be right under
8 GB.)

> |------------------+--------------+---------------+----------------------|
> | Performance      |              | High          | Need to target areas |
> |                  |              |               | of the greatest      |
> |                  |              |               | potential for        |
> |                  |              |               | performance          |
> |                  |              |               | improvement.  Some   |
> |                  |              |               | suggestions include: |
> |                  |              |               | Better warm cache    |
> |                  |              |               | performance and      |
> |                  |              |               | better write         |
> |                  |              |               | performance.         |
> |------------------+--------------+---------------+----------------------|

How about doing something about the horrendous vos-dump performance.
Surely that can be made faster. Even if it means a 'local only' vos dump
that has to be run on the server itself.

> |------------------+--------------+---------------+----------------------|
> | Win NT Mount (so | In Progress  | N/A           | Paul Hill  (MIT)     |
> | you don't need   |              |               |                      |
> | /machinename/afs)|              |               |                      |
> |------------------+--------------+---------------+----------------------|

Having "/LOCALHOST-AFS" would even be an extreme improvement on this.

> |------------------+--------------+---------------+----------------------|
> | Multiple back up |              | Low (large    |                      |
> | volume instances |              | effort        |                      |
> |                  |              | relative to   |                      |
> |                  |              | return)       |                      |
> |------------------+--------------+---------------+----------------------|

I actually did some looking on this and what I noticed is that it didn't
look like implementing multiple volume instances would be a problem at
all (looked downright easy in fact) - however, it looked like _keeping
track of them_ would be the problem. 

Some things on my agenda as soon as I get our new test cell in place and
running (we finally got hardware to do the test cell):

	better log file handling - adding these flags:
		-syslog: log to syslogs directly
		-onelog: for salvager only - log to a single log file, prefixing lines
with
			volname, or partition, depending on how called
		-datelog: all servers - name the log files with a date when they are
first opened
	
	options to turn off certain logging - in particular, that annoying
WhoAmI failure caused by NT clients

	
The above should be simple... some other stuff I'd like to do at some
point

	structured output of vos commands - i.e. listvol and listvldb - output
in a tabular or XML style structure to make it easy to parse with common
libraries

	possibly implement a server-side 'voldump' command (similar in use to
volinfo) that would emulate 'vos dump' but would go directly against
local disk for efficiency.



-- Nathan

------------------------------------------------------------
Nathan Neulinger                       EMail:  nneul@umr.edu
University of Missouri - Rolla         Phone: (573) 341-4841
CIS - Systems Programming                Fax: (573) 341-4216