Comments on: What is Multi Version Concurrency Control (MVCC)

By: Vlad Mihalcea

Vlad Mihalcea — Wed, 01 Mar 2017 12:40:51 +0000

I also wrote an article about how MVCC works in PostgreSQL. Ir provides diagrams for how INSERT, UPDATE, and DELETE statements work in the context of MVCC:

https://vladmihalcea.com/2017/03/01/how-does-mvcc-multi-version-concurrency-control-work/

By: Anvesh Patel

Anvesh Patel — Sat, 25 Feb 2017 08:17:57 +0000

In reply to Venkatesh Natukula. You cannot see old data, but you can get a count of dead tuples/old data so that it helps you to take decision for cleaning. You can execute your cleaning command on each table level, schema level and also at a database level.

By: Venkatesh Natukula

Venkatesh Natukula — Sat, 25 Feb 2017 01:42:48 +0000

Thanks to write such a good use full article. I have doubt can we able to see old data by writing any command and if we write any cleaning command is this specific to one table or it will work entire DB. can please share commands regarding these things.

By: daniel

daniel — Tue, 14 Feb 2017 02:55:13 +0000

thank you for amazing fantastic postingg!!

By: Anvesh Patel

Anvesh Patel — Sun, 18 Dec 2016 13:19:33 +0000

In reply to valtih1978.

Really very Interesting point,
Your point is valid where insert with the same RTS will create deadlock situation, but what I have found is this is very rare situation and even you can control your insert using different Isolation level.
Still, I need to research on this and will let you know in detail.

By: valtih1978

valtih1978 — Thu, 01 Dec 2016 19:06:40 +0000

Wikipedia says “If transaction Ti wants to Write to object P, and there is also another transaction Tk happening to the same object, the Read Timestamp RTS(Ti) must precede the Read Timestamp RTS(Tk), i.e., RTS(Ti) < RTS(Tk), for the object Write Operation (WTS) to succeed."

That is, only single user can update anything in the database. For if we have two transactions, none has RTS lower than the other, no matter which DB item we look at. None can write anything. This seems strange. Moreover, they violate this rule in their example when they show that long-running T2 reads values written by T3. Can you explain the crazy rule?

By: Anvesh Patel

Anvesh Patel — Sun, 30 Oct 2016 06:49:56 +0000

In reply to Mauro. Really very nice Question ! Yes, Index will also change pointer from Old version to New version data.

By: Mauro

Mauro — Fri, 28 Oct 2016 13:39:14 +0000

Do you know how indexes are maintained when an update occurs? Does the index remain pointing to the old version of the row and this address points to the new version? Or the index is updated as part of the update process?