Amanda-Users

Re: adding tapes to cycle

2009-05-06 14:43:36
Subject: Re: adding tapes to cycle
From: Chris Hoogendyk <hoogendyk AT bio.umass DOT edu>
To: Brandon Metcalf <brandon AT geronimoalloys DOT com>
Date: Wed, 06 May 2009 11:19:28 -0400


Brandon Metcalf wrote:
----- Chris Hoogendyk <hoogendyk AT bio.umass DOT edu> wrote:
Chris Hoogendyk wrote:
Brandon Metcalf wrote:
I am currently using the following default settings:

  dumpcycle 4 weeks
  runspercycle 20
  tapecycle 25 tapes

If wanted to, say, have an 8 week dumpcycle could I simply add and label 25 tapes and change the parameters above to

  dumpcycle 8 weeks
  runspercycle 40
  tapecycle 50 tapes

and expect Amanda to do the right thing?
Short answer: Yes. Of course.

A bit more: You can change your configuration and then feed it a few more tapes at a time until you reach your full number of tapes. As long as it has a new labeled tape each time it wants one, it will continue to work. When you run out of new labeled tapes, then there need to be enough tapes that there will be an old one that can be re-used. Typically, you have a few more tapes than specified in tapecycle. If runtapes is greater than 1, then you need more -- see: http://wiki.zmanda.com/index.php/Tapecycle.
I want to amend my comments with one caveat.

While Amanda will do "the right thing," to some extent it will be hamstrung by your configuration. You will typically only have one full backup at a time (within a dumpcycle). You will have the constant threat of having a full for some DLE overwritten, and the incrementals that relate to that full becoming meaningless, before you have another full. In addition, a single tape failure can result in multiple DLE's not being recoverable.

If you changed your original scheme to

   dumpcycle 2 weeks
   runspercycle 10
   tapecycle 25 tapes

Then you would always have a full backup, and often two.

If you add 25 tapes and make your new scheme

   dumpcycle 4 weeks
   runspercycle 20
   tapecycle 50 tapes

Then, again, you will always have a full backup.

I don't know if your intention in stretching out the tapecycle is to have more coverage, or whether you have too much data for your network/server/tapes to handle in the shorter cycle. I would be concerned about redundancy. If you aren't having trouble with too much data, I would use a scheme like

   dumpcycle 1 week
   runspercycle 5
   tapecycle 50 tapes

or, perhaps, 2, 10 and 50.

My motivation behind asking the question was to have more coverage.  I 
certainly don't have a problem with too much data as I'm using virtual tapes on 
a large NAS RAID array.  I was thinking that by increasing the number of tapes 
and the dumpcycle and runspercycle parameters, I could recover further back in 
time.  It sounds like this may not be the case which is probably due to my lack 
of understand of exactly how Amanda uses these parameters in its algorithms.

Assuming that I don't have a issue with too much data and I would like the 
following to happen

  * a full backup once per week
  * backups 6 days per week
  * the ability to go back a couple of months

Would a config like the following work?

  dumpcycle 1 week
  runspercycle 6
  tapecycle 75

And could I simply add all of the virtual tapes and once and change the above 
parameters and be OK?

Thanks for the help.

That would work.

You get at least one full backup of each DLE during each dumpcycle. On other runs of amdump during the dumpcycle, it will do incrementals of those DLE's. So every run you get a mix of fulls and incrementals, which evens out the load.

A tapecycle of 50 (which you originally mentioned) would take you out to 8 weeks and a couple of days, given that you are doing 6 runs per week.

A tapecycle of 75 would take you out to 12 and a half weeks.

Watch your NAS carefully. In a sense, you are putting all your eggs in one basket. If it is made up of consumer grade SATA disks, I would even use raid 6 rather than raid 5, and I would still watch the logs religiously to catch any signs of disk failures and swap them out promptly.


--
---------------

Chris Hoogendyk

-
  O__  ---- Systems Administrator
 c/ /'_ --- Biology & Geology Departments
(*) \(*) -- 140 Morrill Science Center
~~~~~~~~~~ - University of Massachusetts, Amherst
<hoogendyk AT bio.umass DOT edu>

---------------
Erdös 4



<Prev in Thread] Current Thread [Next in Thread>