On AIX 5.3, there have been some problems getting PostgreSQL to compile and run using GCC.

You will want to use a version of GCC subsequent to 3.3.2, particularly if you use a prepackaged version. We had good success with 4.0.1. Problems with earlier versions seem to have more to do with the way IBM packaged GCC than with actual issues with GCC, so that if you compile GCC yourself, you might well have success with an earlier version of GCC.

AIX 5.3 has a problem where sockaddr_storage is not defined to be large enough. In version 5.3, IBM increased the size of sockaddr_un, the address structure for Unix-domain sockets, but did not correspondingly increase the size of sockaddr_storage. The result of this is that attempts to use Unix-domain sockets with PostgreSQL lead to libpq overflowing the data structure. TCP/IP connections work OK, but not Unix-domain sockets, which prevents the regression tests from working.

The problem was reported to IBM, and is recorded as bug report PMR29657. If you upgrade to maintenance level 5300-03 or later, that will include this fix. A quick workaround is to alter _SS_MAXSIZE to 1025 in /usr/include/sys/socket.h. In either case, recompile PostgreSQL once you have the corrected header file.

PostgreSQL relies on the system's getaddrinfo function to parse IP addresses in listen_addresses, pg_hba.conf, etc. Older versions of AIX have assorted bugs in this function. If you have problems related to these settings, updating to the appropriate AIX fix level shown above should take care of it.

One user reports:

When implementing PostgreSQL version 8.1 on AIX 5.3, we periodically ran into problems where the statistics collector would “mysteriously” not come up successfully. This appears to be the result of unexpected behavior in the IPv6 implementation. It looks like PostgreSQL and IPv6 do not play very well together on AIX 5.3.

Any of the following actions “fix” the problem.

AIX can be somewhat peculiar with regards to the way it does memory management. You can have a server with many multiples of gigabytes of RAM free, but still get out of memory or address space errors when running applications. One example is loading of extensions failing with unusual errors. For example, running as the owner of the PostgreSQL installation:

=# CREATE EXTENSION plperl;
ERROR:  could not load library "/opt/dbs/pgsql/lib/plperl.so": A memory address is not in the address space for the process.

Running as a non-owner in the group possessing the PostgreSQL installation:

=# CREATE EXTENSION plperl;
ERROR:  could not load library "/opt/dbs/pgsql/lib/plperl.so": Bad address

Another example is out of memory errors in the PostgreSQL server logs, with every memory allocation near or greater than 256 MB failing.

The overall cause of all these problems is the default bittedness and memory model used by the server process. By default, all binaries built on AIX are 32-bit. This does not depend upon hardware type or kernel in use. These 32-bit processes are limited to 4 GB of memory laid out in 256 MB segments using one of a few models. The default allows for less than 256 MB in the heap as it shares a single segment with the stack.

In the case of the plperl example, above, check your umask and the permissions of the binaries in your PostgreSQL installation. The binaries involved in that example were 32-bit and installed as mode 750 instead of 755. Due to the permissions being set in this fashion, only the owner or a member of the possessing group can load the library. Since it isn't world-readable, the loader places the object into the process' heap instead of the shared library segments where it would otherwise be placed.

The “ideal” solution for this is to use a 64-bit build of PostgreSQL, but that is not always practical, because systems with 32-bit processors can build, but not run, 64-bit binaries.

If a 32-bit binary is desired, set LDR_CNTRL to MAXDATA=0xn0000000, where 1 <= n <= 8, before starting the PostgreSQL server, and try different values and postgresql.conf settings to find a configuration that works satisfactorily. This use of LDR_CNTRL tells AIX that you want the server to have MAXDATA bytes set aside for the heap, allocated in 256 MB segments. When you find a workable configuration, ldedit can be used to modify the binaries so that they default to using the desired heap size. PostgreSQL can also be rebuilt, passing configure LDFLAGS="-Wl,-bmaxdata:0xn0000000" to achieve the same effect.

For a 64-bit build, set OBJECT_MODE to 64 and pass CC="gcc -maix64" and LDFLAGS="-Wl,-bbigtoc" to configure. (Options for xlc might differ.) If you omit the export of OBJECT_MODE, your build may fail with linker errors. When OBJECT_MODE is set, it tells AIX's build utilities such as ar, as, and ld what type of objects to default to handling.

By default, overcommit of paging space can happen. While we have not seen this occur, AIX will kill processes when it runs out of memory and the overcommit is accessed. The closest to this that we have seen is fork failing because the system decided that there was not enough memory for another process. Like many other parts of AIX, the paging space allocation method and out-of-memory kill is configurable on a system- or process-wide basis if this becomes a problem.

If PostgreSQL on Windows crashes, it has the ability to generate minidumps that can be used to track down the cause for the crash, similar to core dumps on Unix. These dumps can be read using the Windows Debugger Tools or using Visual Studio. To enable the generation of dumps on Windows, create a subdirectory named crashdumps inside the cluster data directory. The dumps will then be written into this directory with a unique name based on the identifier of the crashing process and the current time of the crash.

You can build with either GCC or Sun's compiler suite. For better code optimization, Sun's compiler is strongly recommended on the SPARC architecture. We have heard reports of problems when using GCC 2.95.1; GCC 2.95.3 or later is recommended. If you are using Sun's compiler, be careful not to select /usr/ucb/cc; use /opt/SUNWspro/bin/cc.

You can download Sun Studio from https://www.oracle.com/technetwork/server-storage/solarisstudio/downloads/. Many of GNU tools are integrated into Solaris 10, or they are present on the Solaris companion CD. If you like packages for older version of Solaris, you can find these tools at http://www.sunfreeware.com. If you prefer sources, look at https://www.gnu.org/prep/ftp.

If configure complains about a failed test program, this is probably a case of the run-time linker being unable to find some library, probably libz, libreadline or some other non-standard library such as libssl. To point it to the right location, set the LDFLAGS environment variable on the configure command line, e.g.,

configure ... LDFLAGS="-R /usr/sfw/lib:/opt/sfw/lib:/usr/local/lib"

See the ld man page for more information.

On Solaris 7 and older, the 64-bit version of libc has a buggy vsnprintf routine, which leads to erratic core dumps in PostgreSQL. The simplest known workaround is to force PostgreSQL to use its own version of vsnprintf rather than the library copy. To do this, after you run configure edit a file produced by configure: In src/Makefile.global, change the line

LIBOBJS =

to read

LIBOBJS = snprintf.o

(There might be other files already listed in this variable. Order does not matter.) Then build as usual.

On the SPARC architecture, Sun Studio is strongly recommended for compilation. Try using the -xO5 optimization flag to generate significantly faster binaries. Do not use any flags that modify behavior of floating-point operations and errno processing (e.g., -fast). These flags could raise some nonstandard PostgreSQL behavior for example in the date/time computing.

If you do not have a reason to use 64-bit binaries on SPARC, prefer the 32-bit version. The 64-bit operations are slower and 64-bit binaries are slower than the 32-bit variants. And on other hand, 32-bit code on the AMD64 CPU family is not native, and that is why 32-bit code is significant slower on this CPU family.

Yes, using DTrace is possible. See Section 28.5 for further information.

If you see the linking of the postgres executable abort with an error message like:

Undefined                       first referenced
 symbol                             in file
AbortTransaction                    utils/probes.o
CommitTransaction                   utils/probes.o
ld: fatal: Symbol referencing errors. No output written to postgres
collect2: ld returned 1 exit status
make: *** [postgres] Error 1

your DTrace installation is too old to handle probes in static functions. You need Solaris 10u4 or newer.