Обсуждение: wal_sender_delay (WalSndDelay) has served its purpose
AFAICS we could get rid of WalSndDelay: there is no longer any reason for the walsender loop to wake up unless it's received a latch event. (Its WaitLatch call is missing WL_POSTMASTER_DEATH right now, but that is easily fixed.) Is anyone sufficiently attached to that GUC to not want to see it go away? regards, tom lane
On Wed, Aug 10, 2011 at 10:23 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote: > AFAICS we could get rid of WalSndDelay: there is no longer any reason > for the walsender loop to wake up unless it's received a latch event. > (Its WaitLatch call is missing WL_POSTMASTER_DEATH right now, but that > is easily fixed.) Is anyone sufficiently attached to that GUC to not > want to see it go away? Please remove. -- Simon Riggs http://www.2ndQuadrant.com/ PostgreSQL Development, 24x7 Support, Training & Services
On Wed, Aug 10, 2011 at 5:29 PM, Simon Riggs <simon@2ndquadrant.com> wrote: > On Wed, Aug 10, 2011 at 10:23 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote: > >> AFAICS we could get rid of WalSndDelay: there is no longer any reason >> for the walsender loop to wake up unless it's received a latch event. >> (Its WaitLatch call is missing WL_POSTMASTER_DEATH right now, but that >> is easily fixed.) Is anyone sufficiently attached to that GUC to not >> want to see it go away? > > Please remove. +1! -- Robert Haas EnterpriseDB: http://www.enterprisedb.com The Enterprise PostgreSQL Company