Thread: Upgrade problem

I've just upgraded from 7.2 to 8.1.5 under AIX 4.3.
gcc 2.95.3
gmake 3.79.1


I ran into a build problem which I was able to resolve by adding adding
'LDFLAGS= -xlinker -bbigtoc' to the configure command line.
make check passes all 98 tests, but when I try to start postgresql I get:

LOG:  could not translate service "5432" to address: Host not found
WARNING:  could not create listen socket for "*"
FATAL:  could not create any TCP/IP sockets

in the logfile.

I reread the INSTALL doc and sec 23.4 of the v8.1 manual and see no
mention of any config changes required for the newer release.

A google search didn't yield much except that it seems that the error
message is generated in src/backend/libpq/pqcomm.c. Can anyone tell what
I've done wrong?

Bill Kurland

Bill Kurland <bill@panix.com> writes:
> I've just upgraded from 7.2 to 8.1.5 under AIX 4.3.
> make check passes all 98 tests, but when I try to start postgresql I get:

> LOG:  could not translate service "5432" to address: Host not found

Hmm, this seems to be another one of those "AIX's getaddrinfo() is
broken" problems :-(.  We patched around one problem last month:
http://developer.postgresql.org/cvsweb.cgi/pgsql/src/backend/libpq/ip.c.diff?r1=1.32.2.1&r2=1.32.2.3

I'm afraid that that patch does not fix your problem, but it would be
worth trying it before we dig any deeper.

            regards, tom lane

Version 1.8 of VTD-XML is now released. The new features are:

·        XMLModifier is a easy to use class that takes advantage of the
incremental update capability offered by VTD-XML

·        XPath built-in functions are now almost complete

·        This release added encoding support for iso-8859-2~10, windows code
page 1250~1258

·        Added various functions to autoPilot that evaluate XPath to string,
number and boolean

·        This release also fixes a number of XPath bugs related to string
handling

To download the latest release please visit http://vtd-xml.sf.net

Tom:

Thanks for the reply. I've installed the patch and, as you predicted, it had no effect.

I did a google search on AIX + getaddrinfo and found

http://lists.samba.org/archive/rsync/2002-April/002063.html

In that context the author says that adding the port number in etc/services solved his problem with getaddrinfo. So I tried that and, lo, it has some effect, though I'm not sure it's 100% desirable. The log entry is:

LOG:  could not bind IPv6 socket: The type of socket is not supported in this protocol family.
HINT:  Is another postmaster already running on port 5432? If not, wait a few seconds and retry.
LOG:  database system was shut down at 2006-11-26 23:26:32 EST
LOG:  checkpoint record is at 0/38D198
LOG:  redo record is at 0/38D198; undo record is at 0/0; shutdown TRUE
LOG:  next transaction ID: 565; next OID: 10793
LOG:  next MultiXactId: 1; next MultiXactOffset: 0
LOG:  database system is ready
LOG:  transaction ID wrap limit is 2147484146, limited by database "postgres"

but postmaster is running.

I've restored my database and a cursory look seems to show everything is working as expected. Can you tell me what the log entries signify?

Thanks again.



Tom Lane wrote:

Bill Kurland <bill@panix.com> writes: 

I've just upgraded from 7.2 to 8.1.5 under AIX 4.3.
make check passes all 98 tests, but when I try to start postgresql I get:   

LOG:  could not translate service "5432" to address: Host not found   

Hmm, this seems to be another one of those "AIX's getaddrinfo() is
broken" problems :-(.  We patched around one problem last month:
http://developer.postgresql.org/cvsweb.cgi/pgsql/src/backend/libpq/ip.c.diff?r1=1.32.2.1&r2=1.32.2.3

I'm afraid that that patch does not fix your problem, but it would be
worth trying it before we dig any deeper.
		regards, tom lane

---------------------------(end of broadcast)---------------------------
TIP 9: In versions below 8.0, the planner will ignore your desire to      choose an index scan if your joining column's datatypes do not      match
 

-- 

Bill Kurland
Shakespeare & Co.
138 Watts Street
New York, NY 10013

212-965-9683
212-965-9684 Fax 

--
Can I now congratulate the same nation upon its freedom? Is it because liberty in the abstract may be classed amongst the blessings of mankind, that I am seriously to felicitate a madman, who has escaped from the protecting restraint and wholesome darkness of his cell, on his restoration to the enjoyment of light and liberty? Am I to congratulate a highwayman and murderer who has broke prison upon the recovery of his natural rights?-Edmund Burke, "Reflections on the Revolution in France" 

Bill Kurland <bill@shakespeare-nyc.com> writes:
> I did a google search on AIX + getaddrinfo and found
> http://lists.samba.org/archive/rsync/2002-April/002063.html
> In that context the author says that adding the port number in
> etc/services solved his problem with getaddrinfo.

Interesting.  I wonder whether IBM thinks that there is some
security-related reason for only allowing programs to bind to port
numbers that are listed in /etc/services?

> So I tried that and,
> lo, it has some effect, though I'm not sure it's 100% desirable. The log
> entry is:

> LOG:  could not bind IPv6 socket: The type of socket is not supported in
> this protocol family.
> HINT:  Is another postmaster already running on port 5432? If not, wait
> a few seconds and retry.

I think this is OK.  There are many machines on which the userspace
code supports IPv6 while the kernel doesn't, or vice versa.  It looks
to me like getaddrinfo returned both IPv4 and IPv6 translations of
"localhost", but the kernel rejected the IPv6 version when PG tried it.
Since you evidently have a working IPv4 port, there's nothing to worry
about.  If it really bugs you, the /etc/netsvc.conf change suggested in
our FAQ_AIX would probably suppress the log message.

Can anyone else confirm the behavior of getaddrinfo wanting port 5432
to be listed in /etc/services?  If this is real, we ought to have
something about it in FAQ_AIX.

            regards, tom lane

tgl@sss.pgh.pa.us (Tom Lane) writes:
> Bill Kurland <bill@shakespeare-nyc.com> writes:
>> I did a google search on AIX + getaddrinfo and found
>> http://lists.samba.org/archive/rsync/2002-April/002063.html
>> In that context the author says that adding the port number in
>> etc/services solved his problem with getaddrinfo.
>
> Interesting.  I wonder whether IBM thinks that there is some
> security-related reason for only allowing programs to bind to port
> numbers that are listed in /etc/services?
>
>> So I tried that and,
>> lo, it has some effect, though I'm not sure it's 100% desirable. The log
>> entry is:
>
>> LOG:  could not bind IPv6 socket: The type of socket is not supported in
>> this protocol family.
>> HINT:  Is another postmaster already running on port 5432? If not, wait
>> a few seconds and retry.
>
> I think this is OK.  There are many machines on which the userspace
> code supports IPv6 while the kernel doesn't, or vice versa.  It looks
> to me like getaddrinfo returned both IPv4 and IPv6 translations of
> "localhost", but the kernel rejected the IPv6 version when PG tried it.
> Since you evidently have a working IPv4 port, there's nothing to worry
> about.  If it really bugs you, the /etc/netsvc.conf change suggested in
> our FAQ_AIX would probably suppress the log message.
>
> Can anyone else confirm the behavior of getaddrinfo wanting port 5432
> to be listed in /etc/services?  If this is real, we ought to have
> something about it in FAQ_AIX.

That doesn't make a lot of sense to me...

It is fair to say that AIX does include a PostgreSQL entry, by
default:

pgorg@nj-570-db4011:/opt/home/pgorg $ grep 5432 /etc/services
postgresql              5432/tcp                # PostgreSQL Database
postgresql              5432/udp                # PostgreSQL Database

[We didn't add these entries ourselves; they were there by default.]

But most of our database backends run on ports other than 5432, and I
haven't noticed anything that seems to tie to that.  We have had
difficulties with getaddrinfo(), but I don't think there's anything
that hasn't either been resolved in modern PG releases or listed in
FAQ_AIX.
--
let name="cbbrowne" and tld="acm.org" in name ^ "@" ^ tld;;
http://linuxfinances.info/info/linux.html
Rules of the Evil Overlord #131. "I will never place the key to a cell
just out of a prisoner's reach." <http://www.eviloverlord.com/>

Chris Browne wrote:
> tgl@sss.pgh.pa.us (Tom Lane) writes:
>
>> Bill Kurland <bill@shakespeare-nyc.com> writes:
>>
>>> I did a google search on AIX + getaddrinfo and found
>>> http://lists.samba.org/archive/rsync/2002-April/002063.html
>>> In that context the author says that adding the port number in
>>> etc/services solved his problem with getaddrinfo.
>>>
>> Interesting.  I wonder whether IBM thinks that there is some
>> security-related reason for only allowing programs to bind to port
>> numbers that are listed in /etc/services?
>>
>>
>>> So I tried that and,
>>> lo, it has some effect, though I'm not sure it's 100% desirable. The log
>>> entry is:
>>>
>>> LOG:  could not bind IPv6 socket: The type of socket is not supported in
>>> this protocol family.
>>> HINT:  Is another postmaster already running on port 5432? If not, wait
>>> a few seconds and retry.
>>>
>> I think this is OK.  There are many machines on which the userspace
>> code supports IPv6 while the kernel doesn't, or vice versa.  It looks
>> to me like getaddrinfo returned both IPv4 and IPv6 translations of
>> "localhost", but the kernel rejected the IPv6 version when PG tried it.
>> Since you evidently have a working IPv4 port, there's nothing to worry
>> about.  If it really bugs you, the /etc/netsvc.conf change suggested in
>> our FAQ_AIX would probably suppress the log message.
>>
>> Can anyone else confirm the behavior of getaddrinfo wanting port 5432
>> to be listed in /etc/services?  If this is real, we ought to have
>> something about it in FAQ_AIX.
>>
>
> That doesn't make a lot of sense to me...
>
> It is fair to say that AIX does include a PostgreSQL entry, by
> default:
>
> pgorg@nj-570-db4011:/opt/home/pgorg $ grep 5432 /etc/services
> postgresql              5432/tcp                # PostgreSQL Database
> postgresql              5432/udp                # PostgreSQL Database
>
> [We didn't add these entries ourselves; they were there by default.]
>
Odd. I checked 4 different AIX 4.3 machines and none of them have
postgresql entries. I don't have
any AIX 5 boxes. Is that what you are running?
> But most of our database backends run on ports other than 5432, and I
> haven't noticed anything that seems to tie to that.  We have had
> difficulties with getaddrinfo(), but I don't think there's anything
> that hasn't either been resolved in modern PG releases or listed in
> FAQ_AIX.
>
I just reread the FAQ_AIX doc for 8.1.5 and I couldn't find any
reference to getaddrinfo. What difficulties
have you seen and under which AIX ?

bill@panix.com (Bill Kurland) writes:
> Chris Browne wrote:
>> tgl@sss.pgh.pa.us (Tom Lane) writes:
>>
>>> Bill Kurland <bill@shakespeare-nyc.com> writes:
>>>
>>>> I did a google search on AIX + getaddrinfo and found
>>>> http://lists.samba.org/archive/rsync/2002-April/002063.html
>>>> In that context the author says that adding the port number in
>>>> etc/services solved his problem with getaddrinfo.
>>>>
>>> Interesting.  I wonder whether IBM thinks that there is some
>>> security-related reason for only allowing programs to bind to port
>>> numbers that are listed in /etc/services?
>>>
>>>
>>>> So I tried that and, lo, it has some effect, though I'm not sure
>>>> it's 100% desirable. The log entry is:
>>>>       LOG:  could not bind IPv6 socket: The type of socket is not
>>>> supported in this protocol family.
>>>> HINT:  Is another postmaster already running on port 5432? If not,
>>>> wait a few seconds and retry.
>>>>
>>> I think this is OK.  There are many machines on which the userspace
>>> code supports IPv6 while the kernel doesn't, or vice versa.  It looks
>>> to me like getaddrinfo returned both IPv4 and IPv6 translations of
>>> "localhost", but the kernel rejected the IPv6 version when PG tried it.
>>> Since you evidently have a working IPv4 port, there's nothing to worry
>>> about.  If it really bugs you, the /etc/netsvc.conf change suggested in
>>> our FAQ_AIX would probably suppress the log message.
>>>
>>> Can anyone else confirm the behavior of getaddrinfo wanting port 5432
>>> to be listed in /etc/services?  If this is real, we ought to have
>>> something about it in FAQ_AIX.
>>>
>>
>> That doesn't make a lot of sense to me...
>>
>> It is fair to say that AIX does include a PostgreSQL entry, by
>> default:
>>
>> pgorg@nj-570-db4011:/opt/home/pgorg $ grep 5432 /etc/services
>> postgresql              5432/tcp                # PostgreSQL Database
>> postgresql              5432/udp                # PostgreSQL Database
>>
>> [We didn't add these entries ourselves; they were there by default.]
>>
> Odd. I checked 4 different AIX 4.3 machines and none of them have
> postgresql entries. I don't have
> any AIX 5 boxes. Is that what you are running?

Yup, we're running 5.3.  I don't think we have any 5.1 boxes left to
look at.

You might also look for port 3306; we've got MySQL listed as
associated with that port.  I suspect AIX 4.3 doesn't include that
either.

I'd be very unkeen on continuing to run 4.anything; we had a nasty
surprise this spring when we discovered that 5.1 was falling out of
service shortly before we were to migrate databases to 5.3.

>> But most of our database backends run on ports other than 5432, and I
>> haven't noticed anything that seems to tie to that.  We have had
>> difficulties with getaddrinfo(), but I don't think there's anything
>> that hasn't either been resolved in modern PG releases or listed in
>> FAQ_AIX.

> I just reread the FAQ_AIX doc for 8.1.5 and I couldn't find any
> reference to getaddrinfo. What difficulties have you seen and under
> which AIX ?

It has always turned into patches.  For 7.4, we needed to apply a
patch to turn parts of the IPv6 functionality off, otherwise we had
intermittent SIG 11 problems.  Nothing directly FAQ-worthy; it was
code that got added into later versions.
--
let name="cbbrowne" and tld="linuxfinances.info" in String.concat "@" [name;tld];;
http://linuxdatabases.info/info/advocacy.html
Oh,  boy, virtual memory!  Now I'm  gonna make  myself a  really *big*
RAMdisk!