How to write a code to timetravel using a specific transaction ID - snowflake-cloud-data-platform

I would like to use a timetravel feature on snowflake and restore the original table.
I've deleted and created the table using following command:
DROP TABLE "SOCIAL_LIVE"
CREATE TABLE "SOCIAL_LIVE" (...)
I would like to go back to the original table before dropping table.
I've used following code (hid the transaction ID to 'xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx')
Select "BW"."PUBLIC"."SOCIAL_LIVE".* From "BW"."PUBLIC"."SOCIAL_LIVE";
select * from SOCIAL_LIVE before(statement => 'xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx');
Received an error message:
Statement xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx cannot be used to specify time for time travel query.
How can we go back to the original table and restore it on snowflake?

The documentation states:
After dropping a table, creating a table with the same name creates a
new version of the table. The dropped version of the previous table
can still be restored using the following method:
Rename the current version of the table to a different name.
Use the UNDROP TABLE command to restore the previous version.
If you need further information, this page is useful:
https://docs.snowflake.net/manuals/sql-reference/sql/drop-table.html#usage-notes
You will need to undrop the table in order to access that data, though. Time-travel is not maintained by name alone. So, once you dropped and recreated the table, the new table has its own, new time travel.

Looks like there's 3 common reasons that error is seen, with solutions:
the table has been dropped and recreated
see this answer
the time travel period has been exceeded
no solution: target a statement within the time travel period for the table
the wrong statement type is being targeted
only certain statement types can be targeted. Currently, these include SELECT, BEGIN, COMMIT, and DML (INSERT, UPDATE etc). See documentation here.

Statement xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx cannot be used to specify time for time travel query.
Usually we will get above error when we trying to travel behind the object creation time. Try with time travel option with offset option.

Related

Tracking changes in SQL Server on a BCP Imported Table

I upload an Excel file using BCP. (Truncate the current table in DB every day and BCP in from the excel file to repopulate table). It is important for me to keep a log of all the changes made to the rows (could be row additions or changes in columns of current rows). The idea is to keep a log of all the changes made.
I have read a few articles online, where we can create a log table and trigger (have no idea how to do it). A table of logs that has columns like
Date | Field | Old Value | New Value.
Firstly, how to do this?
Secondly, whats a smarter way to not log truncating of table and just the actual changes. I'm thinking of creating a temp table (tbl_Excefile_Temp) where I will import the file and then UPDATE the current table (tbl_Excefile) from the tbl_Excefile_Temp This way all the changes made in the current table will get logged automatically in the logs table.
I know its a big use case, could you please guide.
If you are using SQL server 2016 or higher I would advise you to look into temporal tables. If you stop truncating and use a merge statement you have a very easy way of keeping a log. Whenever you make a change SQL server will write to old values away and add the datetimes when the old row was valid.
With temporal tables you can query your table as they were at a specific datetime. In regular use there is no difference with a non-temporal table.

Updating data in Clickhouse

I went over the documentation for Clickhouse and I did not see the option to UPDATE nor DELETE. It seems to me its an append only system.
Is there a possibility to update existing records or is there some workaround like truncating a partition that has records in it that have changed and then re-insering the entire data for that partition?
Through Alter query in clickhouse we can able to delete/update the rows in a table.
For delete: Query should be constructed as
ALTER TABLE testing.Employee DELETE WHERE Emp_Name='user4';
For Update: Query should be constructed as
ALTER TABLE testing.employee UPDATE AssignedUser='sunil' where AssignedUser='sunny';
UPDATE: This answer is no longer true, look at https://stackoverflow.com/a/55298764/3583139
ClickHouse doesn't support real UPDATE/DELETE.
But there are few possible workarounds:
Trying to organize data in a way, that is need not to be updated.
You could write log of update events to a table, and then calculate reports from that log. So, instead of updating existing records, you append new records to a table.
Using table engine that do data transformation in background during merges. For example, (rather specific) CollapsingMergeTree table engine:
https://clickhouse.yandex/reference_en.html#CollapsingMergeTree
Also there are ReplacingMergeTree table engine (not documented yet, you could find example in tests: https://github.com/yandex/ClickHouse/blob/master/dbms/tests/queries/0_stateless/00325_replacing_merge_tree.sql)
Drawback is that you don't know, when background merge will be done, and will it ever be done.
Also look at samdoj's answer.
You can drop and create new tables, but depending on their size this might be very time consuming. You could do something like this:
For deletion, something like this could work.
INSERT INTO tableTemp SELECT * from table1 WHERE rowID != #targetRowID;
DROP table1;
INSERT INTO table1 SELECT * from tableTemp;
Similarly, to update a row, you could first delete it in this manner, and then add it.
Functionality to UPDATE or DELETE data has been added in recent ClickHouse releases, but its expensive batch operation which can't be performed too frequently.
See https://clickhouse.yandex/docs/en/query_language/alter/#mutations for more details.
It's an old question, but updates are now supported in Clickhouse. Note it's not recommended to do many small changes for performance reasons. But it is possible.
Syntax:
ALTER TABLE [db.]table UPDATE column1 = expr1 [, ...] WHERE filter_expr
Clickhouse UPDATE documentation

SQL Server: log modifications

I have a SQL Server and I need to log any changes made to a set of tables and their fields. Information needed is the user, the date time, the related table / field and the new value.
I saw the Change Data Capture (CDC) feature which seems perfect but it requires a non-standard version - and I have (and I may only have) the standard version.
The single solution I see is to use trigger, but it may cause performance troubles (it blocks the related table while the log is being inserted). Is there any other solution?
If you don't want triggers to do this, then define a stored procedure that will insert entries in log, after inserting the data into the table successfully.
it blocks the related table while the log is being inserted
Most probably you are using FOR INSERT for trigger, i think you should try AFTER INSERT

What are good strategies for updating a live database table?

I have a db table that gets entirely re-populated with fresh data periodically. This data needs to be then pushed into a corresponding live db table, overwriting the previous live data.
As the table size increases, the time required to push the data into the live table also increases, and the app would look like its missing data.
One solution is to push the new data into a live_temp table and then run an SQL RENAME command on this table to rename it as the live table. The rename usually runs in sub-second time. Is this the "right" way to solve this problem?
Are there other strategies or tools to tackle this problem? Thanks.
I don't like messing with schema objects in this way - it can confuse query optimizers and I have no idea what will happen to any transactions that are going on while you execute the rename.
I much prefer to add a version column to the table, and have a separate table to hold the current version.
That way, the client code becomes
select *
from myTable t,
myTable_currentVersion tcv
where t.versionID = tcv.CurrentVersion
This also keeps history around - which may or not be useful; if it's not delete old records after setting the CurrentVersion column.
Create a duplicate table - exact copy.
Create a new table that does nothing more than keep track of the "up to date" table.
MostCurrent (table)
id (column) - holds name of table holding the "up to date" data.
When repopulating, populate the older table and update MostCurrent.id to reflect this table.
Now, in your app where you bind the data to the page, bind the newest table.
Would it be appropriate to only push changes to the live db table? For most applications I have worked with changes have been minimal. You should be able to apply all the changes in a single transaction. Committing the transaction will make them visible with no outage on the table.
If the data does change entirely, then you could configure the database so that you can replace all the data in a single transaction.

Deleting Rows from a SQL Table marked for Replication

I erroneously delete all the rows from a MS SQL 2000 table that is used in merge replication (the table is on the publisher). I then compounded the issue by using a DTS operation to retrieve the rows from a backup database and repopulate the table.
This has created the following issue:
The delete operation marked the rows for deletion on the clients but the DTS operation bypasses the replication triggers so the imported rows are not marked for insertion on the subscribers. In effect the subscribers lose the data although it is on the publisher.
So I thought "no worries" I will just delete the rows again and then add them correctly via an insert statement and they will then be marked for insertion on the subscribers.
This is my problem:
I cannot delete the DTSed rows because I get a "Cannot insert duplicate key row in object 'MSmerge_tombstone' with unique index 'uc1MSmerge_tombstone'." error. What I would like to do is somehow delete the rows from the table bypassing the merge replication trigger. Is this possible? I don't want to remove and redo the replication because the subscribers are 50+ windows mobile devices.
Edit: I have tried the Truncate Table command. This gives the following error "Cannot truncate table xxxx because it is published for replication"
Have you tried truncating the table?
You may have to truncate the table and reset the ID field back to 0 if you need the inserted rows to have the same ID. If not, just truncate and it should be fine.
You also could look into temporarily dropping the unique index and adding it back when you're done.
Look into sp_mergedummyupdate
Would creating a second table be an option? You could create a second table, populate it with the needed data, add the constraints/indexes, then drop the first table and rename your second table. This should give you the data with the right keys...and it should all consist of SQL statements that are allowed to trickle down the replication. It just isn't probably the best on performance...and definitely would impose some risk.
I haven't tried this first hand in a replicated environment...but it may be at least worth trying out.
Thanks for the tips...I eventually found a solution:
I deleted the merge delete trigger from the table
Deleted the DTSed rows
Recreated the merge delete trigger
Added my rows correctly using an insert statement.
I was a little worried bout fiddling with the merge triggers but every thing appears to be working correctly.

Resources