Re: Initial 9.2 pgbench write results

----- Цитат от Robert Haas (robertmhaas@gmail.com), на 28.02.2012 в 19:25 ----- <br /><br />> On Tue, Feb 28, 2012
at11:36 AM, Jeff Janes  wrote: <br />>> How hard would it be to dummy up a bgwriter which, every time it wakes
<br/>>> up, it forks off a child process to actually do the write, and then <br />>> the real one just
waitsfor the child to exit?  If it didn't have to <br />>> correctly handle signals, SINVAL, and such, it should
bejust a few <br />>> lines of code, but I don't know how much we can ignore signals and <br />>> such even
justfor testing purposes.  My thought here is that the <br />>> kernel is getting in a snit over one process
doingall the writing on <br />>> the system, and is punishing that process in a way that ruins things <br
/>>>for everyone. <br />> <br />> I would assume the only punishment that the kernel would inflict would
<br/>> be to put the bgwriter to sleep. That would make the bgwriter less <br />> effective, of course, but it
shouldn'tmake it worse than no bgwriter <br />> at all. Unless it does it some really stupid way, like making <br
/>>bgwriter sleep while it holds some lock. <br />> <br />> But maybe I'm missing something - what do you have
inmind? <br />> <br />> -- <br />> Robert Haas <br />> EnterpriseDB: http://www.enterprisedb.com <br />>
TheEnterprise PostgreSQL Company <br />> <br />> -- <br />> Sent via pgsql-hackers mailing list
(pgsql-hackers@postgresql.org)<br />> To make changes to your subscription: <br />>
http://www.postgresql.org/mailpref/pgsql-hackers<br />> <br />> <br /><br />-- <br />Luben Karavelov