I was a little surprised to discover that DDL statements (alter table, create index etc) implicitly commit the current transaction in MySQL. Coming from MS SQL Server, the ability to do database alterations in a transaction locally (that was then rolled back) was an important part of my workflow. For continuous integration, the rollback was used if the migration hiccuped for any reason, so that at least we did not leave the database in a half-migrated state.
How do people solve these two problems when using MySQL with migrations and continuous integration?
DDL statements cause an implicit commit and there is nothing you can do about it. There is no way to stop this behaviour.
Which DDL statements have this behaviour changes over time so you need to check for your version.
5.1 http://dev.mysql.com/doc/refman/5.1/en/implicit-commit.html
5.5 http://dev.mysql.com/doc/refman/5.5/en/implicit-commit.html
5.6 http://dev.mysql.com/doc/refman/5.6/en/implicit-commit.html
When we are just extending the schema, new tables/columns/views/procs/etc, that will not affect existing code then automation is OK, just check for errors and fix them.
When they will affect existing code then you need to devise a strategy on a case by case basis. Since there is no rollback you need your own backout plan and you need to test it thoroughly.
Since it is case-by-case there is not a lot that I can offer in the way of help for your particular situation.
One possibility is doing DDL changes in a non-destructive-manner, which would include:
split logic in DDL/DCL (+1 to reverse all) and DML
run only the DDL/DCL script adding columns, new tables, ..
depending on result:
on success, apply the DML changes,
on fail, apply reverse DDL/DCL script removing the stuff you wanted to add in second step (obviously with some errors "does not exist" depending on how far step 1 got)
remove what is not needed anymore, drop old columns/tables
Related
I am planning to use pt-online-schema-change tool for a table alter on a production server in a replication environment. Wanted to understand what steps are executed when pt-online-schema-change tool aborts due to server tool. Is it possible to resume after it aborts due to server load ? will it drop the temp table and the triggers it created? Will i need to start all over again?
pt-online-schema-change is very verbose with exactly what it's doing. So, when something fails I've always been able to read the last lines and it'll tell me what to do.
Specifically when migrations have failed due to load, triggers have not been dropped. In that case the output has stated exactly what I need to execute to drop them. It's possible that they are kept to be able to resume. I don't know about that. In my case I've always dropped the triggers and started from scratch after tweaking parameters or lowering database usage. I guess my way of doing it works if you aren't scripting pt-online-schema-change executions.
Regarding automating dropping temp table and triggers; By default it drops them (see documentation). However, you can always specify --nodrop-old-table, --nodrop-new-table and --nodrop-triggers. For larger tables (when I have the disk space) I generally always specify --nodrop-old-table. This enables me to quickly roll back to the old table in case something goes wrong when it's swapped in by simply issuing
RENAME TABLE mytable TO mytable_failedmigration, old-table TO mytable;
If the migration succeeds I drop my giant table by following the instructions in https://serverfault.com/a/566710/37237.
I made a wrong update query in my table.
I forgot to make an id field in the WHERE clause.
So that updated all my rows.
How to recover that?
I didn't have a backup....
There are two lessons to be learned here:
Backup data
Perform UPDATE/DELETE statements within a transaction, so you can use ROLLBACK if things don't go as planned
Being aware of the transaction (autocommit, explicit and implicit) handling for your database can save you from having to restore data from a backup.
Transactions control data manipulation statement(s) to ensure they are atomic. Being "atomic" means the transaction either occurs, or it does not. The only way to signal the completion of the transaction to database is by using either a COMMIT or ROLLBACK statement (per ANSI-92, which sadly did not include syntax for creating/beginning a transaction so it is vendor specific). COMMIT applies the changes (if any) made within the transaction. ROLLBACK disregards whatever actions took place within the transaction - highly desirable when an UPDATE/DELETE statement does something unintended.
Typically individual DML (Insert, Update, Delete) statements are performed in an autocommit transaction - they are committed as soon as the statement successfully completes. Which means there's no opportunity to roll back the database to the state prior to the statement having been run in cases like yours. When something goes wrong, the only restoration option available is to reconstruct the data from a backup (providing one exists). In MySQL, autocommit is on by default for InnoDB - MyISAM doesn't support transactions. It can be disabled by using:
SET autocommit = 0
An explicit transaction is when statement(s) are wrapped within an explicitly defined transaction code block - for MySQL, that's START TRANSACTION. It also requires an explicitly made COMMIT or ROLLBACK statement at the end of the transaction. Nested transactions is beyond the scope of this topic.
Implicit transactions are slightly different from explicit ones. Implicit transactions do not require explicity defining a transaction. However, like explicit transactions they require a COMMIT or ROLLBACK statement to be supplied.
Conclusion
Explicit transactions are the most ideal solution - they require a statement, COMMIT or ROLLBACK, to finalize the transaction, and what is happening is clearly stated for others to read should there be a need. Implicit transactions are OK if working with the database interactively, but COMMIT statements should only be specified once results have been tested & thoroughly determined to be valid.
That means you should use:
SET autocommit = 0;
START TRANSACTION;
UPDATE ...;
...and only use COMMIT; when the results are correct.
That said, UPDATE and DELETE statements typically only return the number of rows affected, not specific details. Convert such statements into SELECT statements & review the results to ensure correctness prior to attempting the UPDATE/DELETE statement.
Addendum
DDL (Data Definition Language) statements are automatically committed - they do not require a COMMIT statement. IE: Table, index, stored procedure, database, and view creation or alteration statements.
Sorry man, but the chances of restoring an overwritten MySQL database are usually close to zero. Different from deleting a file, overwriting a record actually and physically overwrites the existing data in most cases.
To be prepared if anything comes up here, you should stop your MySQL server, and make a copy of the physical directory containing the database so nothing can get overwritten further: A simple copy+paste of the data folder to a different location should do.
But don't get your hopes up - I think there's nothing that can be done really.
You may want to set up a frequent database backup for the future. There are many solutions around; one of the simplest, most reliable and easiest to automate (using at or cron in Linux, or the task scheduler in Windows) is MySQL's own mysqldump.
Sorry to say that, but there is no way to restore the old field values without a backup.
Don't shoot the messenger...
Do you have binlogs enabled? You can recover by accessing the binlogs.
I wrote a tool for our project, for applying sql update files that were committed, to the DB. Whenever run (on deployment), it calculates the list of update files which need to be applied, and applies them iniside a transaction.
Recently I became aware of an issue: mysql would implicitly commit a transaction, whenever DDL statements (like create) are executed. http://dev.mysql.com/doc/refman/5.0/en/implicit-commit.html
This is an issue for me, as sometimes an sql update file contains several statements, which as I understand will result in committing the transaction in the middle of executing the update file. This is a problem, because whenever a subsequent update will fail (which happens from time to time) I want to be able to rollback the transaction, or at least track which update files where applied (completely) and which were not.
Is there a way around the implicit transactions issue? I.e. is there a way to rollback a sequence of DDL statements whenever one of them fail?
Any other suggestions how I can handle the issue?
Thanks
Gidi
No. MySQL does not support transactional DDL. You either need to separate your DDL statements from DML statements, or perhaps try to use migration tool like RuckUsing
I made a wrong update query in my table.
I forgot to make an id field in the WHERE clause.
So that updated all my rows.
How to recover that?
I didn't have a backup....
There are two lessons to be learned here:
Backup data
Perform UPDATE/DELETE statements within a transaction, so you can use ROLLBACK if things don't go as planned
Being aware of the transaction (autocommit, explicit and implicit) handling for your database can save you from having to restore data from a backup.
Transactions control data manipulation statement(s) to ensure they are atomic. Being "atomic" means the transaction either occurs, or it does not. The only way to signal the completion of the transaction to database is by using either a COMMIT or ROLLBACK statement (per ANSI-92, which sadly did not include syntax for creating/beginning a transaction so it is vendor specific). COMMIT applies the changes (if any) made within the transaction. ROLLBACK disregards whatever actions took place within the transaction - highly desirable when an UPDATE/DELETE statement does something unintended.
Typically individual DML (Insert, Update, Delete) statements are performed in an autocommit transaction - they are committed as soon as the statement successfully completes. Which means there's no opportunity to roll back the database to the state prior to the statement having been run in cases like yours. When something goes wrong, the only restoration option available is to reconstruct the data from a backup (providing one exists). In MySQL, autocommit is on by default for InnoDB - MyISAM doesn't support transactions. It can be disabled by using:
SET autocommit = 0
An explicit transaction is when statement(s) are wrapped within an explicitly defined transaction code block - for MySQL, that's START TRANSACTION. It also requires an explicitly made COMMIT or ROLLBACK statement at the end of the transaction. Nested transactions is beyond the scope of this topic.
Implicit transactions are slightly different from explicit ones. Implicit transactions do not require explicity defining a transaction. However, like explicit transactions they require a COMMIT or ROLLBACK statement to be supplied.
Conclusion
Explicit transactions are the most ideal solution - they require a statement, COMMIT or ROLLBACK, to finalize the transaction, and what is happening is clearly stated for others to read should there be a need. Implicit transactions are OK if working with the database interactively, but COMMIT statements should only be specified once results have been tested & thoroughly determined to be valid.
That means you should use:
SET autocommit = 0;
START TRANSACTION;
UPDATE ...;
...and only use COMMIT; when the results are correct.
That said, UPDATE and DELETE statements typically only return the number of rows affected, not specific details. Convert such statements into SELECT statements & review the results to ensure correctness prior to attempting the UPDATE/DELETE statement.
Addendum
DDL (Data Definition Language) statements are automatically committed - they do not require a COMMIT statement. IE: Table, index, stored procedure, database, and view creation or alteration statements.
Sorry man, but the chances of restoring an overwritten MySQL database are usually close to zero. Different from deleting a file, overwriting a record actually and physically overwrites the existing data in most cases.
To be prepared if anything comes up here, you should stop your MySQL server, and make a copy of the physical directory containing the database so nothing can get overwritten further: A simple copy+paste of the data folder to a different location should do.
But don't get your hopes up - I think there's nothing that can be done really.
You may want to set up a frequent database backup for the future. There are many solutions around; one of the simplest, most reliable and easiest to automate (using at or cron in Linux, or the task scheduler in Windows) is MySQL's own mysqldump.
Sorry to say that, but there is no way to restore the old field values without a backup.
Don't shoot the messenger...
Do you have binlogs enabled? You can recover by accessing the binlogs.
I'm being given a data source weekly that I'm going to parse and put into a database. The data will not change much from week to week, but I should be updating the database on a regular basis. Besides this weekly update, the data is static.
For now rebuilding the entire database isn't a problem, but eventually this database will be live and people could be querying the database while I'm rebuilding it. The amount of data isn't small (couple hundred megabytes), so it won't load that instantaneously, and personally I want a bit more of a foolproof system than "I hope no one queries while the database is in disarray."
I've thought of a few different ways of solving this problem, and was wondering what the best method would be. Here's my ideas so far:
Instead of replacing entire tables, query for the difference between my current database and what I want to place in the database. This seems like it could be an unnecessary amount of work, though.
Creating dummy data tables, then doing a table rename (or having the server code point towards the new data tables).
Just telling users that the site is going through maintenance and put the system offline for a few minutes. (This is not preferable for obvious reasons, but if it's far and away the best answer I'm willing to accept that.)
Thoughts?
I can't speak for MySQL, but PostgreSQL has transactional DDL. This is a wonderful feature, and means that your second option, loading new data into a dummy table and then executing a table rename, should work great. If you want to replace the table foo with foo_new, you only have to load the new data into foo_new and run a script to do the rename. This script should execute in its own transaction, so if something about the rename goes bad, both foo and foo_new will be left untouched when it rolls back.
The main problem with that approach is that it can get a little messy to handle foreign keys from other tables that key on foo. But at least you're guaranteed that your data will remain consistent.
A better approach in the long term, I think, is just to perform the updates on the data directly (your first option). Once again, you can stick all the updating in a single transaction, so you're guaranteed all-or-nothing semantics. Even better would be online updates, just updating the data directly as new information becomes available. This may not be an option for you if you need the results of someone else's batch job, but if you can do it, it's the best option.
BEGIN;
DELETE FROM TABLE;
INSERT INTO TABLE;
COMMIT;
Users will see the changeover instantly when you hit commit. Any queries started before the commit will run on the old data, anything afterwards will run on the new data. The database will actually clear the old table once the last user is done with it. Because everything is "static" (you're the only one who ever changes it, and only once a week), you don't have to worry about any lock issues or timeouts. For MySQL, this depends on InnoDB. PostgreSQL does it, and SQL Server calls it "snapshotting," and I can't remember the details off the top of my head since I rarely use the thing.
If you Google "transaction isolation" + the name of whatever database you're using, you'll find appropriate information.
We solved this problem by using PostgreSQL's table inheritance/constraints mechanism.
You create a trigger that auto-creates sub-tables partitioned based on a date field.
This article was the source I used.
Which database server are you using? SQL 2005 and above provides a locking method called "Snapshot". It allows you to open a transaction, do all of your updates, and then commit, all while users of the database continue to view the pre-transaction data. Normally, your transaction would lock your tables and block their queries, but snapshot locking would be perfect in your case.
More info here: http://blogs.msdn.com/craigfr/archive/2007/05/16/serializable-vs-snapshot-isolation-level.aspx
But it requires SQL Server, so if you're using something else....
Several database systems (since you didn't specify yours, I'll keep this general) do offer the SQL:2003 Standard statement called MERGE which will basically allow you to
insert new rows into a target table from a source which don't exist there yet
update existing rows in the target table based on new values from the source
optionally even delete rows from the target that don't show up in the import table anymore
SQL Server 2008 is the first Microsoft offering to have this statement - check out more here, here or here.
Other database system probably will have similar implementations - it's a SQL:2003 Standard statement after all.
Marc
Use different table names(mytable_[yyyy]_[wk]) and a view for providing you with a constant name(mytable). Once a new table is completely imported update your view so that it uses that table.