HOME

5h-3LL 1.0
DIR: /usr/share/doc/dirvish
/usr/share/doc/dirvish/
Current File : /usr/share/doc/dirvish/TODO.html
<HTML>
<HEAD>
<TITLE>Dirvish ToDo List</TITLE>
<META NAME="ID" CONTENT="$Id: TODO.html,v 12.0 2004/02/25 02:42:14 jw Exp $">
<META NAME="Version tag" CONTENT="$Name: Dirvish-1_2 $">
</HEAD>
<BODY>
<H1 align="center">Dirvish ToDo List</H1>
<P>
List items are not in a prioritised order.
<P>
List item marks designate some intent.
<UL>
<LI type="square">to be done before next version.
<LI type="disc">normal.
<LI type="circle">ongoing.
</UL>
<h3>dirvish</h3>
<ul>
<LI type=square>
Allow vault[:branch] as an argument so --vault and --branch
options aren't needed.
<LI type="disc">
Enhance the <b>password-file:</b> parameter 
so that if given a relative
path use the config-file search function (when written).
<LI type="disc">
Add a --retry option for retrying a failed dirvish image.
<p>
This would require pre-existance of destination image.
<p>
Moves summary, log and rsync_error files into try.#
directories.
<LI type="disc">
Add a <b>meta-owner</b> parameter like <b>meta-perm</b>
to specify a owner:group for meta-files in images.
<LI type="circle">
Enhance recognition and assessment of rsync errors and warnings.
</ul>
<h3>dirvish-expire</h3>
<ul>
<LI type=square>
Allow vault[:branch] as an argument so --vault
option isn't needed.
<LI type="disc">
Move non-expiration notices to bottom of dirvish-expire output.
<LI type="disc">
Put functionality for doing cleanups of
hist files to deal with deleted backups somewhere in dirvish-expire.
A possible approach would be for
<B>--histck</B> to remove hist entries having no matching images on disk.
<B>--hist</B> might remove hist entry during expiration but
that would complicate things.
<LI type="disc">
Rework to eliminate the internal find.
Have dirvish-expire walk the vaults like dirvish-locate.
<LI type="disc">
Improve internal logic to handle failed
images better. Instead of using just sort order it should
track a latest successful image to keep.  Preferring an
unexpired, but saving an expired one if no unexpired
available.
</ul>
<h3>dirvish-locate</h3>
<ul>
<LI type="disc">
Improve the performance of index file parsing.
<LI type="disc">
Report all file dates with year even those without year in
index.
</ul>
<h3>dirvish-runall</h3>
<ul>
<li type="disc">
Allow an optional timespec like expire-rule to make runall
stansas conditional.
</ul>
<h3>in general</h3>
<ul>
<li type=square>
Allow config-file search that searches
<I>bank</i>/<i>vault</i>/<b>dirvish</b>/ also search
/<b>etc</b>/<b>dirvish</b>/<i>vault</i>/ for files.
<br>
Do this with a new function that can also be called for
other file-based looking.
<li type="disc">
Add command-line options for non-interactive install.sh runs.
<li type="disc">
Add the ability to create an initial set of configs to install.sh.
</ul>
<hr>
<h1 align="center">Dirvish Wish List</h1>
<p>
These items are low-priority.
It is recognised that they have merit
and they have been given some thought but are, at present,
not considered sufficiently important to invest the needed
time.
If a paying client wished one of them the priority would change
but until that happens or the author's requirements change they are shelved.
<UL>
<LI>
Add a mechanism for dirvish and dirvish-expire to log their
activity.
<p>
This would probably a config parameter where you could
either specify a file name or a syslog "facility" such as
daemon or local3.
<LI>
Create a <b>dirvish-sync</b> utility that could syncronize
with a remote vault.
<p>
Useful for secondary backup servers and removable hard
drives used for offsites.
<p>
This would examine the images in the specified vault.
Remove those images from the destination that are not in the source.
For each new image in creation order
rsync the whole image
with the best available reference image for the link-dest.
Finish by rsyncing the dirvish directory if any.
<LI>
Add option to run arbitrary command instead of rsync.
<P>
This would allow dirvish to schedule jobs that don't
involve rsync but belong under the same process management as
the main backups.
Having to maintain multiple backup systems and schedulers
can be a pain.
<LI>
Extend EXPIRE-RULES to support multi-stage actions
<P>
The date specifier of an expire rule would contain a
designator set elsewhere in the config file.  The value of
the designator (expire action) would be a list of date-specs
and actions to be done.
<p>For example:
<pre>
expire-rule
	wd { sun } md { 1-7 }               expire-monthly

expire-monthly
	+1 month	clean-tree clean_list
	+6 months	delete-tree
	+1 year		delete-image
</pre>
<p>
The actual format will probably differ and the image manager
(dirvish-expire) will need to know what to do with the actions
and record completion in the summary file to prevent repeats.
<p>
Probable actions would be:
<table>
<tr valign="top"><td>delete-image</td>
	<td>delete whole image including metadata.</td></tr>
<tr valign="top"><td>delete-tree</td>
	<td>delete image tree leaving metadata.</td></tr>
<tr valign="top"><td>compress-tree</td>
	<td>
create a compressed archive of the tree (tree.cpio.gz)
and delete the tree.
	</td></tr>
<tr valign="top"><td>clean-tree</td>
	<td>
lean the image tree by deleting files that match patterns in a file.
	</td></tr>
<tr valign="top"><td>run</td>
	<td>run an executable/script.</td></tr>
</table>
<LI>
Add <B>--replace</B> and <B>--inplace</B> to config to allow overwriting of an existing image.
<P>
If reference == image don't
delete tree and don't use <B>--link-dest</B>.
<P>
If reference != image and <B>--replace</B>, delete image before reuse 
because the delta from the reference image is most likely
less than the reused image
But if <B>--inplace</B> specified don't delete even if
reference != image.  Inplace is especially useful if the
backup is incomplete and you wish to re-run it manually.
<LI>
Create the scheduler.
<P>
Most of the functionality is already
documented in dirvish.conf(5) as comments not visible to man.
<P>
The scheduler would generate a priority ordered run queue of
jobs that can be run and launch as many would be permitted
by the Max-jobs and Max-load limits.
When a job completes recalculate whether any jobs still on the queue can run
(window and limits) and launch them.
If the queue is empty rescan for jobs that may now be runnable.
Exit when there are no running or runnable jobs.
<P>
The scheduler would use a /var/run/dirvish.pid file to
prevent multiple dirvish-sched processes from running at
once.
The scheduler could be launched by cron at intervals.
<LI>
Create a file security like facility
<p>
This would examine the xfer log and
report changed files after running them through a filter
using include/exclude patterns or rules in a config file to eliminate
expected changes.  Excluded files (ones expected to change)
might still be reported if they are suid/sgid.
<p>
Deletes would not be detected this way because dirvish
doesn't see them.
</UL>
</BODY>
</HTML>