search

denshoproject / ddr-local

Web UI used for interacting with DDR collections and entities on a local machine.
Other
3 stars 0 forks source link

Existing objects being overwritten #197

Closed gjost closed 8 years ago

gjost commented 8 years ago

Micah:

This morning when I fired up DDR to add new objects to an existing collection I was getting an error screen when creating the new object, but after refreshing the page it started overwriting previous objects. I tried to sync the collection (after re-entering the correct object) and the sync failed. Could it be an issue with the new server not knowing these objects already exist? I've attached screen shots below.

Froh:

FYI I pulled the latest DDR code with the newer version of git-annex on Friday(?) of last week. The make install script ran without complaints and I re-added the pubkey for mits to /etc/hosts. Things seemed to be working from the command-line and from the webui.

Note: This is the master branch. Debian 7/wheezy.

Micah:

So- last week on Thursday Geoff updated Workbench. We contacted him when we tried to use it for the first time since December and got an error message when trying to open our collections. Everything seemed fine- but that day I didn't try and create any new records. Today was my first day back since last Thursday, and when I went to create a new record in ONE-2 (the Shiuko Sakai Collection) I got an error message. I backed out and checked to see if any records were created- none were. I tried to create a new record again- I got the error message again and pushed the refresh button and the error cleared and blank record was created- BUT it wasn't a new record. The record was number 2-265 which I had already created last year. I repopulated the fields with the record information and re-attached the photos. Then I tried to make a new record again, and it did the same thing, expect it erased 2-266. I repopulated the fields and attached photos and emailed Geoff (erasing 2-267 to make screen shot of it). I tried to sync the system to see if that would correct the error (since these records were created just before you all brought the new server online) and that sync failed- twice. Then I tried saving all the records that might be affected and tried to create a new record again. This time erasing record 2-268. Took a break since we hadn't heard back from Geoff yet, and discussed what was happening with Todd. We decided to push forward through the error and repopulate the remaining 9 objects as I tried to create new records. Once I made it through the records we had already created that were being over written new records were being created.

gjost commented 8 years ago

master branch, last Thursday (2016-02-18), these commits:

gjost commented 8 years ago

Froh:

not seeing anything weird on the git-annex branch. I manually pulled (as ddr) origin master, then pushed origin master, then switched btranched to git-annex. Then pulled origin git-annex, pushed origin git-annex, switched back to master branch, pulled and pushed. It was reporting "143 commits ahead of master" before I manually pulled/pushed.

gjost commented 8 years ago

Looked at Cgit and tried to piece together the sequence of events. Note: Cgit not the greatest at history browsing, the pages use offsets so links to pages of commits won't be useful for long.

This is an odd series of commits (in reverse chrono order):

8f4db44a1c  2016-02-24 11:01:26 (PST)  Added entity file(s)
    Note: Another file is added.

77377a0d04  2016-02-24 11:00:29 (PST)  Added entity file(s)
    Note: File is added.

2d1772162b  2016-02-24 10:58:59 (PST)  Updated entity file(s)
    Note: Previous values are replaced.

115438e419  2016-02-24 10:55:16 (PST)  Updated entity file(s)
    Note: One second after previous commit.  Several fields are updated.

0b253a862d  2016-02-24 10:55:15 (PST)  Initialized entity ddr-one-2-265
    Note: Previously populated entity.json is overwritten with empty one(!).  A new blank file should be added here!

This pattern is repeated with 2-266.

Looking forward a page in the Cgit commit logs, it looks like the sequence of commits starting with 0b253a862d was replayed on top of files created on 2015-12-10, starting with this commit:

21560a2011  2015-12-10 11:12:53 (PST)  Initialized entity ddr-one-2-265
gjost commented 8 years ago

Froh:

OK -- looks like it was because they added entities ddr-one-2-265 - ddr-one-2-277 and then did not sync before we did the mits --> mits2 swapover. I had emailed them, but it must have slipped through the cracks. She was also worried about ddr-one-5, but I checked it and it appears that they had synced that one, so everything appears copacetic. Emergency averted -- nothing to see here....