Re: txid and current_timestamp - Mailing list pgsql-general

From Tom Lane
Subject Re: txid and current_timestamp
Date
Msg-id 4508.1340716795@sss.pgh.pa.us
Whole thread Raw
In response to Re: txid and current_timestamp  (Jasen Betts <jasen@xnet.co.nz>)
List pgsql-general
Jasen Betts <jasen@xnet.co.nz> writes:
> On 2012-06-26, Vlad Arkhipov <arhipov@dc.baikal.ru> wrote:
>> Is it guaranteed that if txid2 > txid1 then current_timestamp in
>> transaction 2 >= current_timestamp in transaction 1?

> no.

To enlarge on that: current_timestamp is set at the moment of receipt
from the client of a transaction's first command.  XID is not set until
(and unless) the transaction does something that modifies the database.
The elapsed time between can be quite variable depending on what
commands the client issues.

Even if this weren't the case, I wouldn't recommend relying on such an
assumption, because of factors like clock skew between different
processors.

            regards, tom lane

pgsql-general by date:

Previous
From: Stuart Bishop
Date:
Subject: Re: Minimal streaming replication
Next
From: "Albe Laurenz"
Date:
Subject: Re: dblink causing import errors