Opened 7 years ago
Closed 7 years ago
#3904 closed defect (fixed)
PostgreSQL 11 crashing on PostGIS 2.5 trunk
| Reported by: | robe | Owned by: | pramsey |
|---|---|---|---|
| Priority: | blocker | Milestone: | PostGIS 2.5.0 |
| Component: | postgis | Version: | 2.4.x |
| Keywords: | Cc: |
Description
Since this pg11 commit Debbie's pg 11 runs have been failing. I haven't checked PostGIS 2.4 since we don't test 11 on 2.4
Summary
Add pg_noinline macro to c.h. (details)
Improve sys/catcache performance. (details)
Commit a0247e7a11bb9f5fd55694b594a3906b7bd05881 by andres
Add pg_noinline macro to c.h.
Forcing a function not to be inlined can be useful if it's the slow-path
of a performance critical function, or should be visible in profiles to
allow for proper cost attribution.
Author: Andres Freund Discussion:
https://postgr.es/m/20170914061207.zxotvyopetm7lrrp@alap3.anarazel.de
The file was modified src/include/c.h
Commit 141fd1b66ce6e3d10518d66d4008bd368f1505fd by andres
Improve sys/catcache performance.
The following are the individual improvements: 1) Avoidance of
FunctionCallInfo based function calls, replaced by
more efficient functions with a native C argument interface. 2) Don't
extract columns from a cache entry's tuple whenever matching
entries - instead store them as a Datum array. This also allows to
get rid of having to build dummy tuples for negative & list
entries, and of a hack for dealing with cstring vs. text weirdness. 3)
Reorder members of catcache.h struct, so imortant entries are more
likely to be on one cacheline. 4) Allowing the compiler to specialize
critical SearchCatCache for a
specific number of attributes allows to unroll loops and avoid
other nkeys dependant initialization. 5) Only initializing the ScanKey
when necessary, i.e. catcache misses,
greatly reduces cache unnecessary cpu cache misses. 6) Split of the
cache-miss case from the hash lookup, reducing stack
allocations etc in the common case. 7) CatCTup and their corresponding
heaptuple are allocated in one
piece.
This results in making cache lookups themselves roughly three times as
fast - full-system benchmarks obviously improve less than that.
I've also evaluated further techniques:
- replace open coded hash with simplehash - the list walk right now
shows up in profiles. Unfortunately it's not easy to do so safely as
an entry's memory location can change at various times, which
doesn't work well with the refcounting and cache invalidation.
- Cacheline-aligning CatCTup entries - helps some with performance,
but the win isn't big and the code for it is ugly, because the
tuples have to be freed as well.
- add more proper functions, rather than macros for
SearchSysCacheCopyN etc., but right now they don't show up in
profiles.
The reason the macro wrapper for syscache.c/h have to be changed, rather
than just catcache, is that doing otherwise would require exposing the
SysCache array to the outside. That might be a good idea anyway, but
it's for another day.
Author: Andres Freund Reviewed-By: Robert Haas Discussion:
https://postgr.es/m/20170914061207.zxotvyopetm7lrrp@alap3.anarazel.de
The file was modified src/tools/pgindent/typedefs.list
The file was modified src/include/utils/syscache.h
The file was modified src/backend/utils/cache/catcache.c
The file was modified src/include/utils/catcache.h
The file was modified src/backend/utils/cache/syscache.c
The postgis regress log looks like this:
PostgreSQL 11devel on x86_64-pc-linux-gnu, compiled by gcc (Debian 5.2.1-22) 5.2.1 20151010, 64-bit Postgis 2.5.0dev - r15979 - 2017-10-13 22:01:30 scripts 2.5.0dev r15979 GEOS: 3.7.0dev-CAPI-1.11.0 a51a06b1 PROJ: Rel. 4.9.2, 08 September 2015 SFCGAL: 1.3.0 Running tests loader/Point .............. ok loader/PointM .............. ok loader/PointZ .............. ok loader/MultiPoint .............. ok loader/MultiPointM .............. ok loader/MultiPointZ .............. ok loader/Arc .............. ok loader/ArcM .............. ok loader/ArcZ .............. ok loader/Polygon .............. ok loader/PolygonM .............. ok loader/PolygonZ .............. ok loader/TSTPolygon ......... ok loader/TSIPolygon ......... ok loader/TSTIPolygon ......... ok loader/PointWithSchema ..... ok loader/NoTransPoint ......... ok loader/NotReallyMultiPoint ......... ok loader/MultiToSinglePoint ......... ok loader/ReprojectPts ........ ok loader/ReprojectPtsGeog ........ ok loader/Latin1 .... ok loader/Latin1-implicit .... ok loader/mfile .... ok dumper/literalsrid ....... ok dumper/realtable ....... ok affine .. ok bestsrid .. ok binary .. ok boundary .. ok cluster .. ok concave_hull .. ok ctors .. ok curvetoline .. ok dump .. ok dumppoints .. ok empty .. ok estimatedextent .. failed (diff expected obtained: /var/lib/jenkins/workspace/postgis/regress_pgdev/tmp/2_4_pg11.0w64/test_38_diff) ----------------------------------------------------------------------------- --- estimatedextent_expected 2017-01-01 03:28:16.191591396 +0000 +++ /var/lib/jenkins/workspace/postgis/regress_pgdev/tmp/2_4_pg11.0w64/test_38_out 2017-10-13 22:03:16.164629222 +0000 @@ -1,41 +1,4 @@ -WARNING: stats for "t.g" do not exist -#877.1| -WARNING: stats for "t.g" do not exist -#877.2| -WARNING: ST_Estimated_Extent signature was deprecated in 2.1.0. Please use ST_EstimatedExtent -WARNING: stats for "t.g" do not exist -#877.2.deprecated| -WARNING: stats for "t.g" do not exist -#877.3|||| -#877.4|-10.15000|20.15000|-50.40000|30.40000 -#877.5|-10.15000|20.15000|-50.40000|30.40000 -WARNING: stats for "c1.g" do not exist -#3391.1|||| -WARNING: stats for "c2.g" do not exist -#3391.2|||| -WARNING: stats for "p.g" do not exist -#3391.3|||| -#3391.4|0.00|1.00|0.00|1.00 -WARNING: stats for "c2.g" do not exist -#3391.5|||| -#3391.6|0.00|1.00|0.00|1.00 -#3391.7|0.00|1.00|0.00|1.00 -#3391.8|-1.00|0.00|-1.00|0.00 -#3391.9|-1.01|1.01|-1.01|1.01 -#3391.10|0.00|1.00|0.00|1.00 -#3391.11|-1.00|0.00|-1.00|0.00 -#3391.12|-1.01|2.02|-1.01|2.02 -WARNING: stats for "p.g" do not exist -#3391.13|||| -WARNING: stats for "p.g" do not exist -#3391.14|||| -WARNING: stats for "c1.g" do not exist -#3391.15|||| -WARNING: stats for "c1.g" do not exist -#3391.16|||| -#3391.17|0.00|1.00|0.00|1.00 -WARNING: stats for "p.g" do not exist -#3391.18|||| -#3391.19|0.00|1.00|0.00|1.00 -#3391.20|0.00|1.00|0.00|1.00 -NOTICE: drop cascades to 2 other objects +server closed the connection unexpectedly + This probably means the server terminated abnormally + before or while processing the request. +connection to server was lost ----------------------------------------------------------------------------- forcecurve .. failed (diff expected obtained: /var/lib/jenkins/workspace/postgis/regress_pgdev/tmp/2_4_pg11.0w64/test_39_diff) ----------------------------------------------------------------------------- --- forcecurve_expected 2017-01-01 03:28:18.503591901 +0000 +++ /var/lib/jenkins/workspace/postgis/regress_pgdev/tmp/2_4_pg11.0w64/test_39_out 2017-10-13 22:03:16.192628299 +0000 @@ -1,12 +1 @@
Then a lot of the rest is just the pg11 trying to get up and then crashing again
-ndovm1|{1,2,3,4,5,8}
-ndovm2|{1,2,4,6,7}
-#3777|GEOMETRYCOLLECTION EMPTY|1
-#3777|LINESTRING(0 0,0 1)|1
-#3777|POINT EMPTY|1
-#3777|POINT(0 0)|3
-#3777|POINT(0 1)|1
-#3777.1|GEOMETRYCOLLECTION EMPTY|1
-#3777.1|LINESTRING(0 0,0 1)|1
-#3777.1|POINT EMPTY|1
-#3777.1|POINT(0 0)|3
-#3777.1|POINT(0 1)|1
+psql: FATAL: the database system is in recovery mode
-----------------------------------------------------------------------------
orientation .. failed (diff expected obtained: /var/lib/jenkins/workspace/postgis/regress_pgdev/tmp/2_4_pg11.0w64/test_54_diff)
-----------------------------------------------------------------------------
--- orientation_expected 2017-03-17 00:52:28.667011501 +0000
+++ /var/lib/jenkins/workspace/postgis/regress_pgdev/tmp/2_4_pg11.0w64/test_54_out 2017-10-13 22:03:16.276625529 +0000
@@ -1,36 +1 @@
-1|t
-2|t
:
:
-pgcast_06|POLYGON((0 0,0 1,1 1,1 0,0 0))
+psql: FATAL: the database system is in recovery mode
-----------------------------------------------------------------------------
out_geography .. failed (diff expected obtained: /var/lib/jenkins/workspace/postgis/regress_pgdev/tmp/2_4_pg11.0w64/test_56_diff)
-----------------------------------------------------------------------------
--- out_geography_expected 2017-01-01 03:28:16.555591475 +0000
+++ /var/lib/jenkins/workspace/postgis/regress_pgdev/tmp/2_4_pg11.0w64/test_56_out 2017-10-13 22:03:16.284625266 +0000
@@ -1,68 +1 @@
I suspect this might be something that needs fixing upstream so we'll have to come up with an example not involving PostGIS that exercises the mess.
Change History (4)
comment:1 by , 7 years ago
comment:2 by , 7 years ago
So, the crash was coming in a SearchSysCache2(STATRELATT, ...) call.
Which is odd because... STATRELATT hasn't been a syscache key since before PgSQL 9.0 (that's as far as the source I have available to me goes back). The valid key is STATRELATTINH and it's a 3-parameter key. So, this fix might be applicable to all our other branches too, since the 3-parameter STATRELATTINH key goes back as far as any version we support.
Note:
See TracTickets
for help on using tickets.
I've taken 11 out of our 2.5 stream until the issue in 11 is resolved. If it resolves after upstream changes, the nightly PostgreSQL dev git / PostGIS test build will catch it.