I have never worked with Flyway, but all of a sudden I'm managing it. It appears outOfOrder is false by default and I am now working with a failed migration that is blocking continuous deployment to our QA environment. I merged V21__alter_sequence.sql after V22__add_column.sql was already merged. Migrations are all run by a Jenkins build process.
My question is, how do I fix this? My guesses are:
merge U21__alter_sequence.sql then merge V23__alter_sequence.sql?
update flyway.conf, add line flyway.outOfOrder=true and maybe some magic happens on the next merge?
(see my ignorance on this topic!) Any help is greatly appreciated.
Related
I have a question but it turns out I have similar problems to other Github threads that haven't been resolved yet, might as well ask it here.
I am running Django and is using redis-rq to execute background jobs. Unfortunately, it doesn't execute jobs from the default queue. I have already tried to rename it but no progress so far.
Have you guys had this similar event before, and how did you resolve it? Thanks!
We've been using Liquibase for our database migrations, which runs off the Migrations.xml file. This references all of our migrations so it looks something to the effect of:
<include file="ABC.xml" relativeToChangelogFile="true" />
<include file="DEF.xml" relativeToChangelogFile="true" />
<include file="HIJ.xml" relativeToChangelogFile="true" />
...
However, over time, this has grown immensely long. Looking to refactor this has anyone solved this issue? What options are available to essentially condense all these going forward?
Thanks
There is no single answer, what works best depends on your environment, process, and why you want to trim them down.
Generally, I recommend just letting your changelog grow because existing steps are the ones you have tested, deployed, and know are correct. Changing them introduces risk which is always good to avoid. Keeping your changelog unchanged also allows you to continue to bootstrap new dev/QA databases the same way older instances were.
While there is some performance penalty to having a large changelog, Liquibase tries to keep it to a minimum and is primarily just parsing the changelog file and then reading from the DATABASECHANGELOG table which should scale well. Since liquibase update is usually ran infrequently, even if it takes a couple seconds to run it's often not a big deal.
All that being said, there are times it makes sense clear out your changelog. The best way to do that depends on what the problem you are trying to resolve is.
The easiest approach is usually to just remove the include references to your oldest changelog files. If you know that all your databases have the changesSets in ABC.xml and DEF.xml already applied you can just remove the references to them from your master changelog and everthing will be fine. Liquibase doesn't care if there are "unknown" changeSets in the DATABASECHANGELOG table. If you want to continue boostrapping dev and QA environments with ABC.xml and DEF.xml, you can create a second legacy" master changelog.xml which include them and either run both changelogs when needed or make sure the legacy changelog contains both the old and new changelogs.
If you do not want to completely remove changeLog references, you can manually modify the existing changeSets. Often times there are just a few changes that can make a big difference in update performance. For example, if an index was created and then dropped you can skip the cost of the create by removing the drop and create changeSets. Again, liquibase doesn't care for databases that already ran the changeSets and will not see them for new ones. If you have some databases that may still have the index and you want it removed, you can use an indexExists precondition on the drop changeSet after you delete the create changeset.
Precondition checks can be expensive as well, especially depending on the database vendor. Sometimes removing now-unneeded precondition checks can improve performance as well.
So far this has not been an issue for us. Yes, you do end up with a lot of includes but so far I simply did not care. I would strongly suggest not to delete any old changeset or making changes to them. Try to keep everything reproduceable as-is.
Maybe it helps to group changesets by the version of your application. With that I mean that in your migrations.xml file include the files for each of version of your application which include your "real" changesets. That way you could also easily figure out which changeset belongs to which version of your application.
When I go to SpecRunner.html in my browser, the unit tests run fine. The issue I am having is if I change one of the tests, or the code that it tests and refresh the page, it doesn't load the new tests or change at all. I thought that this would be a cache issue, but I have the chrome dev tools disable cache selected.
What am I doing wrong?
Thanks
EDIT: tried restarting my computer, nothing. Clearing cache, nothing. I don't understand why this would be happening.
EDIT2: tried force reload, nothing... changed the file name and reference and still loaded the old code...
This might not answer your specific problem, as there are a million reasons this could happen.
For me, it was because I had deleted some files and then re-added them as a git submodule. After doing this, ls was showing me the OLD files, and I didn't realize that I had to go up a directory and then back in to interact with the new files I had just pulled down.
Old question but I ended up here when looking for an answer, so here's my solution in case it helps anyone - I was having a similar problem with jasmine-rails, removing jasmine-specs.js from public/assets did it for me. I'm using the jasmine-rails gem from https://github.com/searls/jasmine-rails.
This question already exists, but it is over one year old now and a lot has probably happened if the documentation is a good judge. There is no documented path to migrate from current redmine (2.1) to chiliproject for example.
Chiliproject is a fork of redmine, but I am unable to decide wherever I should migrate or not. There is no clear path as to how I should do the migrations and how much functionality I might loose.
Is there a way to compare the differences between the two projects? Is it worth to spend the time investigating the migration path?
If you have migrated what is your experience?
I searched StackOverflow for the "redmine vs. chiliproject" question because I am having a lot of trouble with installing plugins of any kind on the newest chiliproject version.
Usually, it looks like everything is working fine until you try to update the settings for the plugin (for example, install the Contact Form plugin and try to change something on http://SERVER:3000/settings?tab=contact_form), the debug log shows that the changes were made in the database, but they changes are never loaded back to the plugin page.
I have not been ale to find any documentation on potential changes to the plugin architecture in ChiliProject that would cause this. The plugin page does not list many plugins that are known to work with ChiliProject 3 either.
TL;DR: If you think that you will have any desire to use any existing plugins to extend the functionality of the program you choose, go with Redmine.
I just uploaded my first Django project to Heroku and, thus far, it's working will with one exception.
Every time I push an updated version of the project to Heroku...
git push heroku master
...Haystack no longer recognizes any of the items in the database, resulting in empty search results. This isn't even a migration; just a regular push of any updated code. When new items are added, these appear in a search. But when another sync takes place, the results turn up empty. I'm using Haystack with Woosh.
I'm not sure how to rectify this or what may be causing this. If anyone has any experience or knowledge that may help, it would be welcomed.