| From: | Hartmut Holzgraefe <hartmut(dot)holzgraefe(at)gmail(dot)com> |
|---|---|
| To: | pgsql-hackers(at)lists(dot)postgresql(dot)org |
| Subject: | Re: zheap: a new storage format for PostgreSQL |
| Date: | 2018-03-01 16:58:34 |
| Message-ID: | f0e83be7-cd39-3eb4-0c57-4de77d8d0ddc@gmail.com |
| Views: | Raw Message | Whole Thread | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-hackers |
On 01.03.2018 16:30, Satyanarayana Narlapuram wrote:
> Given transaction aborts are expensive, is there any impact on the crash
> recovery?
In InnoDB/XtraDB, which has used the "move old row versions to UNDO log"
since the very beginning, rollbacks are indeed costly, and especially
so on recovery when the UNDO log pages are not yet cached in RAM.
There's is a cost trade of between this kind of "optimistic MVCC" and
rollback/recovery that one has to be aware of.
We get support issues about this at MariaDB every once in a while, but
it is not happening that often.
I can dig up some more info on this from the InnoDB side if you are
interested ...
--
hartmut
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Tom Lane | 2018-03-01 17:04:10 | Re: Missing comment edit |
| Previous Message | Daniel Verite | 2018-03-01 16:57:22 | Re: pgbench - add \if support |