Database marked as different if disconnected

Question

Collaboration server version 1.1, Windows server 2012, Tomcat 8.5.34, elasticsearch 6.2.4

I've created a simple database with reference unit/flow properties data and a single user-made flow. If I make a new repository, commit this database, disconnect from the repository, and then reconnect it, then all parts of the database will be marked as "new". I need to fetch and then commit. Is this the intended behavior? Is there anything I can do short of fetching to get openLCA to recognize that there are actually no changes? Does this happen because when we disconnect the database, openLCA loses information all information about the commits and so has to fetch the entire database just to ensure it's at the right place? Does the fact that all of our version numbers are 0 make this more difficult?

For a single flow, this isn't really that big of a deal, but say we have a much larger database that we share via flash drive or something or even download as a json-ld from collaboration server. The first time we connect this database, we have to fetch the whole thing again. (About to make another post on our fetching issues).

Link to log file from test

Answer 1

Yes this is the intended behaviour. Currently, when disconnecting, the internal index is discarded, and you need to get up to date by fetching. If the data sets are unchanged openLCA should recognize this though and only update the internal index when fetching, without actually downloading and importing any data. Please let me know if this is not the case for you, and how you created the database in that case (When simply disconnecting and reconnecting this is normally correctly recognized by openLCA).

Comments

This is actually a good question. I've added a link to a log file I created using one of our databases. A little about that file. I connected and empty local database to an existing CS database and fetched. I then disconnected from the repository, closed the database, and re-opened openLCA. The log file starts at re-opening openLCA. I should also note that this is just a copy and paste from the console - the log files rotated through too fast to catch the whole thing (the 3 MB limit). I reconnected to the repository and then fetched to get "up to date," which is what makes up most of the log file.

The actual downloading part of the fetch was much quicker (perhaps the downloaded json was cached?"), but then it took about 7 minutes to do whatever it does. The log file looks to me like it's importing data, only much slower because now it's checking the existing data for conflicts.

---

Before actually fetching, does openLCA show a list of differences (Fetch Diff View, where it shows what has changed - the second one, not the one with only the commit list) or is it empty (or mostly empty)?

---

The "Changes to be fetched" window shows the 1 commit - under that commit everything is shown with a +.
The diff window never pops up.

---

Ok both as it should be, but now I don't understand why it imports a lot of data sets. I rechecked the code for v1.1.0 - if no diff is shown it should not fetch anything and thus not import anything (that is also why the downloading is "so fast"). I just recently fixed an issue where partial fetches (when data sets are merged) were incorrectly executed, but this only applies to situations where the diff is shown, conflicts occur and those conflicts are solved by merging the state into something different than the remote state.

During this fix I implemented an additional check in openLCA, so if any merged data set is still fetched it won't be imported again, but this should be unrelated. We will release a new openLCA 1.10.1 version the next days, followed by a collaboration server release 1.2.x soon after. For now I'd say we wait if the issue remains in the new version, and I will do another test on this before the release.