We want to sync from 1 source to 2 different destinations.
Is it possible to store the database file on a different location? or is there another way to get a sync from one source to more than one destination?
change database location
- Posts: 3
- Joined: 26 Feb 2014
- Posts: 3
- Joined: 26 Feb 2014
Can nobody help me to solve this problem?
- Posts: 85
- Joined: 28 Aug 2012
I cannot help you, but I'd like to understand your question.
Two-way sync, right?
1) A-B = new/changed data travels both ways
2) A-C = A gets data from C, which gets data from A (and thus also from B, because of step 1)
3) A changes. If you now run A-B again, B will overwrite changes of C on A (changes that C was not aware of and had not intended, deletions are restored etc). A may not care, but the next time you run A-C, this might create havoc on C, which find its own stuff lost (overwritten/ replaced by B). The same will happen to B of course.
I just fail to understand how you would keep data integrity when it gets altered in 3 different locations.
Two-way sync, right?
1) A-B = new/changed data travels both ways
2) A-C = A gets data from C, which gets data from A (and thus also from B, because of step 1)
3) A changes. If you now run A-B again, B will overwrite changes of C on A (changes that C was not aware of and had not intended, deletions are restored etc). A may not care, but the next time you run A-C, this might create havoc on C, which find its own stuff lost (overwritten/ replaced by B). The same will happen to B of course.
I just fail to understand how you would keep data integrity when it gets altered in 3 different locations.
-
- Site Admin
- Posts: 7211
- Joined: 9 Dec 2007
The are two main synchronization scenarios:
1. mirror: One source, multiple targets are not a problem, just use multiple folder pairs and specify the same folder on the source side (usually left).
2. two-way: Changes on both sides are propagated. An arbitrary number of folders can participate in a "net of synchronized folders". Any change in any folder is propagated to all others as soon as it gets the chance. Therefore a star-like synchronization layout is most useful. That means, each of your clients syncs in two-way mode against one "master" which contains the most recent data. Data integrity is not a problem since all changes are based on multiple syncs, each of which boils down to a single folder pair in two-way mode, one after another.
1. mirror: One source, multiple targets are not a problem, just use multiple folder pairs and specify the same folder on the source side (usually left).
2. two-way: Changes on both sides are propagated. An arbitrary number of folders can participate in a "net of synchronized folders". Any change in any folder is propagated to all others as soon as it gets the chance. Therefore a star-like synchronization layout is most useful. That means, each of your clients syncs in two-way mode against one "master" which contains the most recent data. Data integrity is not a problem since all changes are based on multiple syncs, each of which boils down to a single folder pair in two-way mode, one after another.
-
- Posts: 2450
- Joined: 22 Aug 2012
@Zenju:
Regarding above method 2, two way:
Does this mean the sync has to be run twice.
The first time to assure that the "master" contains all most recent data,
where ever that might originate (at the "master" or at any of the "slaves"),
and then the second run to assure all "slaves" contain all most recent data ?
Or is the upfront compare that smart that it looks beyond plain left-right pair comparison,
effectively first determining what data must first be transferred from any of the "slaves"
to the "master", and only then what must then be transferred from the "master"
to all (or all-but-one) "slaves" ?
Regarding above method 2, two way:
Does this mean the sync has to be run twice.
The first time to assure that the "master" contains all most recent data,
where ever that might originate (at the "master" or at any of the "slaves"),
and then the second run to assure all "slaves" contain all most recent data ?
Or is the upfront compare that smart that it looks beyond plain left-right pair comparison,
effectively first determining what data must first be transferred from any of the "slaves"
to the "master", and only then what must then be transferred from the "master"
to all (or all-but-one) "slaves" ?
- Posts: 85
- Joined: 28 Aug 2012
@Zenju:
Thanks for clarifying. The master-hub-star analogy made me understand. The goal is not to sync together 3 instances, but rather to ALWAYS sync back to the master first. In other words, you don't allow working on B and C simultaneously. C has to wait for B to finish and sync back to the master. Same the other way round. You never really have 3 different versions.
Thanks for clarifying. The master-hub-star analogy made me understand. The goal is not to sync together 3 instances, but rather to ALWAYS sync back to the master first. In other words, you don't allow working on B and C simultaneously. C has to wait for B to finish and sync back to the master. Same the other way round. You never really have 3 different versions.
-
- Site Admin
- Posts: 7211
- Joined: 9 Dec 2007
> Does this mean the sync has to be run twice.
If you have N folders that shall be synced in "two way" mode and one of these folders is changed (either create, update or delete) then it needs to sync against N - 1 folders so that all N folders are in sync again. There's no need to do this immediately depending on how important it is in your scenario to always have all N folders up to date.
"master" and "slave" actually aren't the correct terms for the "two-way" mode since all N folder pairs are equally important in the sense that each change is propagated to all others.
If you have N folders that shall be synced in "two way" mode and one of these folders is changed (either create, update or delete) then it needs to sync against N - 1 folders so that all N folders are in sync again. There's no need to do this immediately depending on how important it is in your scenario to always have all N folders up to date.
"master" and "slave" actually aren't the correct terms for the "two-way" mode since all N folder pairs are equally important in the sense that each change is propagated to all others.
