Koozali.org: home of the SME Server
Obsolete Releases => SME Server 8.x => Topic started by: Jáder on July 23, 2013, 12:08:36 AM
-
I´ve been reading http://wiki.contribs.org/ShadowCopy and I´d like to understand:
1) reading this warning:
"
You will need enough free disk space [and backup capacity] to take a copy of the entire /home/e-smith/files directory plus any changes that take place.
"
why I´ll need +100% of used space if shadowcopy it´s a snapshot and should just copy what was updated AFTER if was enabled.
So if I enable it as 4pm, and run once a day (cronjob at 7pm) just what changed from 4 to 7pm should get copied on 1st snapshot.
2) There are an easy way to do NOT take any snapshot/shadowcopy for directories as default?
3) I´d like to use snapshots like big storages do... just copying changed blocks not your entire FS first time. :(
I know... I should to buy a BIG and EXPENSIVE storage... but I´ll need to stay with SME ... no money at all to spend in storage for my own use! :) I´m just using a 2x 1TB HDDs in RAID1 and moving to a 4x 1TB RAID5 in a few weeks (I hope).
There are any " hack" over there to do this ? I´m not sure if it´s this way it works on VSS on Windows Server.
About this: VSS appears to be a atomic copy (just pointers on FS or something like that), and ShadowCopy as linux implemented appears to be a normal copy. Am I wrong?
Thanks by any tip about where to read/learn about this and by an info / tip from actual user of this on SME.
Regards
Jáder
-
jader
Been a while since I used it.
I think you are wrongly concluding & misinterpreting usage & function.
IIRC the first run of shadow copy will copy all the files in the non excluded ibays & user home folders, in effect everything has changed as there are no files in the shadow copy folders
/home/e-smith/files/.shadow/1
/home/e-smith/files/.shadow/2
& so on
Thus the need to have a lot of spare space on your hard drive that will be at least the size of everything in /home/e-smith/files/
plus spare for subsequent shadow copy runs.
Subsequent runs of shadow copy will backup only what has changed since the first run.
As mentioned external Backup capacity will also need to be sufficiently large to handle all the additional shadow copy files on your server
The Howto advises the commands to exclude specific ibays & users (home folders) from the shadow copy process. I'm not aware of being able to specify specific directories.
AFAIK you always need a reference snapshot in order to track changes, how would you know what has changed.
Simply said, if you want to avail the benefits of shadow copy (ie easy user restore of files & in effect multiple backups throughout the day) then you need to have sufficient hard disk storage. If you cannot provide the required disk space then do not use Shadow Copy.
There were some links to further info at the end of the How to, and I would also search these forums & Google generally for more info.
http://wiki.contribs.org/ShadowCopy
For recovering deleted files use the Recycle Bin feature
http://wiki.contribs.org/RecycleBin
Using the Recycle Bin (on sme server) & a good normal backup system may be sufficient for your needs & avoid the overhead required for Shadow Copy.
Affa can run extra incremental backups throughout the day, to act like a pseudo Shadow Copy, but will need an admin to restore the files.
+1-1 =0
PS: Take look at Comments 9 & 10 of this bug, from when Shadow Copy was being developed, re expected function.
http://bugs.contribs.org/show_bug.cgi?id=1549
-
Janet
Thank by your answer, below some comments:
IIRC the first run of shadow copy will copy all the files in the non excluded ibays & user home folders
(...)
Subsequent runs of shadow copy will backup only what has changed since the first run.
(...)
AFAIK you always need a reference snapshot in order to track changes, how would you know what has changed.
(...)
I removed most of your reply, just let the important topics to anyone undestand what I´ll say:
1) even if the name has COPY on it, it could use simbolic link to avoid WASTE disk space. This is LINUX!!!
2) I don´t think this (copy all content on 1st time) work this way in Microsoft implementation... and is against " the MS way" any linux solution is evaluated.
3) I´d love to help this to be a better solution than MS.
So I´ll look at links (item 9 & 10 of developing thread) and try to find out how to help them to create a better solution!
Regards
Jáder
-
1) even if the name has COPY on it, it could use simbolic link to avoid WASTE disk space. This is LINUX!!!
No - you can't - You can't use symlinks or hardlinks
- You need to copy the file
- check if it has changed, then make a new copy if necessary
if you link to a file you will just edit two files at the same time
2) I don´t think this (copy all content on 1st time) work this way in Microsoft implementation... and is against " the MS way" any linux solution is evaluated.
it doesn't have to be the same as MS - either by pradigm or in practice
3) I´d love to help this to be a better solution than MS.
can only go south if you start this argument
-
I´m sorry... I do not want to be rude. :$
I´d love to see a similar MS solution .
I just think it´s strange to have a second copy of everything just as a starter point :(
It´s works the way it works... so I learned it.
Thanks
Jáder