Re: autovacuum locking question
Jeff Janes <[email protected]> Fri, 6 Dec 2019 19:59:00 -0500
| Newsgroups | gmane.comp.db.postgresql.performance |
|---|---|
| Message-ID | <CAMkU=1xxTrCBUUXM2KtJPa0sk4ZgZUsxHo52SdqVXR-=VU2Dtw@mail.gmail.com> |
--000000000000ad59d5059912ab98 Content-Type: text/plain; charset="UTF-8" On Fri, Dec 6, 2019 at 12:50 PM MichaelDBA <[email protected]> wrote: > And Just to reiterate my own understanding of this... > > autovacuum priority is less than a user-initiated request, so issuing a > manual vacuum (user-initiated request) will not result in being cancelled. > Somethings happen in some situations and not in others. I don't know that it is useful to categorize them into a monotonic priority scale. Autovacs "to prevent wraparound" don't get cancelled the way ordinary autovacs do, but they still use autovac IO throttling settings, not the unthrottled (by default settings) manual vacuum settings, which can be a major problem sometimes. Note that no kind of vacuum should normally get cancelled using the signalling mechanism during truncation phase, that seems to be due to some rather extreme situation with IO congestion. Cheers, Jeff --000000000000ad59d5059912ab98 Content-Type: text/html; charset="UTF-8" Content-Transfer-Encoding: quoted-printable <div dir=3D"ltr"><div dir=3D"ltr">On Fri, Dec 6, 2019 at 12:50 PM MichaelDB= A <<a href=3D"mailto:[email protected]">[email protected]</a>&= gt; wrote:<br></div><div class=3D"gmail_quote"><blockquote class=3D"gmail_q= uote" style=3D"margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,2= 04);padding-left:1ex"> <div>And Just to reiterate my own understanding=20 of this...<br> <br> autovacuum priority is less than a user-initiated request, so issuing a=20 manual vacuum (user-initiated request) will not result in being=20 cancelled.<br></div></blockquote><div><br></div><div>Somethings=C2=A0happen= in some situations and not in others.=C2=A0 I don't know that it is us= eful to categorize them into a monotonic priority scale.</div><div><br></di= v><div>Autovacs "to prevent wraparound" don't get cancelled t= he way ordinary autovacs do, but they still use autovac IO throttling setti= ngs, not the unthrottled (by default settings) manual vacuum settings, whic= h can be a major problem sometimes.</div><div><br></div><div>Note that no k= ind of vacuum should normally get cancelled using the signalling mechanism = during truncation phase, that seems to be due to some rather extreme situat= ion with IO congestion.</div><div><br></div><div>Cheers,</div><div><br></di= v><div>Jeff</div></div></div> --000000000000ad59d5059912ab98--