Обсуждение: AIX and ipv6
we're still having issues with PG on AIX 6.1, our configurations have no
ipv6 as our WAN is purely ipv4, but if we use listen_addresses='*', we
consistently get an error...
LOG: could not bind IPv6 socket: Address already in use
the workaround we've been using is to set listen_addresses to
'127.0.0.1, x.y.z.w' where thats this server's IP. messy, and annoying.
is their any chance this will ever be addressed?
--
john r pierce N 37, W 122
santa cruz ca mid-left coast
John R Pierce <pierce@hogranch.com> writes: > we're still having issues with PG on AIX 6.1, our configurations have no > ipv6 as our WAN is purely ipv4, but if we use listen_addresses='*', we > consistently get an error... > LOG: could not bind IPv6 socket: Address already in use This was discussed at bug #6494 http://archives.postgresql.org/pgsql-bugs/2012-02/msg00190.php It appears to me to be an AIX bug, so you really ought to be pestering IBM not us. Any workaround we might implement could introduce worse problems. regards, tom lane
On 09/11/2012 09:09 AM, John R Pierce wrote: > we're still having issues with PG on AIX 6.1, our configurations have no > ipv6 as our WAN is purely ipv4, but if we use listen_addresses='*', we > consistently get an error... > > LOG: could not bind IPv6 socket: Address already in use > > > the workaround we've been using is to set listen_addresses to > '127.0.0.1, x.y.z.w' where thats this server's IP. messy, and annoying. > > is their any chance this will ever be addressed? Did not say what version you where on. But in 9.1: http://www.postgresql.org/docs/9.1/interactive/runtime-config-connection.html "The special entry * corresponds to all available IP interfaces. The entry 0.0.0.0 allows listening for all IPv4 addresses and :: allows listening for all IPv6 addresses." > > -- Adrian Klaver adrian.klaver@gmail.com
On 09/11/12 10:12 AM, Adrian Klaver wrote: > Did not say what version you where on. But in 9.1: > > http://www.postgresql.org/docs/9.1/interactive/runtime-config-connection.html > "The special entry * corresponds to all available IP interfaces. The entry 0.0.0.0 > allows listening for all IPv4 addresses and :: allows listening for all IPv6 addresses." we've had this problem with 8.4, 9.0 and 9.1. frustratingly, my development server doesn't show the problem, but I may have disabled ipv6 somehow. if it was up to me, we'd toss AIX out and stick with x86 linux for postgresql deployments. but, our overseas production sites like big brand name iron -- john r pierce N 37, W 122 santa cruz ca mid-left coast
On Tue, Sep 11, 2012 at 09:09:22AM -0700, John R Pierce wrote: > > is their any chance this will ever be addressed? The problem you're having is that "*" means "bind to all the addresses on this machine", and for some reason IBM's strange and wonderful implementation of the IP layer appears to give you an IPv6 address from one perspective and not from another. The application (Postgres in this case) can't fix this. Best, A -- Andrew Sullivan ajs@anvilwalrusden.com
On Tue, Sep 11, 2012 at 11:09 AM, John R Pierce <pierce@hogranch.com> wrote: > we're still having issues with PG on AIX 6.1, our configurations have no > ipv6 as our WAN is purely ipv4, but if we use listen_addresses='*', we > consistently get an error... > > LOG: could not bind IPv6 socket: Address already in use > > > the workaround we've been using is to set listen_addresses to '127.0.0.1, > x.y.z.w' where thats this server's IP. messy, and annoying. > > is their any chance this will ever be addressed? have you got your aix completely up to date with all patches? I got burned in a similar fashion a while back on a related call -- getaddrinfo IIRC. merlin
On 09/11/12 12:55 PM, Merlin Moncure wrote: >> is their any chance this will ever be addressed? > have you got your aix completely up to date with all patches? I got > burned in a similar fashion a while back on a related call -- > getaddrinfo IIRC. It was updated to 'current' a year or so ago, I believe. We like to keep our development system at approximately the same TL level as the production systems my development system is currently at... # oslevel -s 6100-06-05-1115 which is AIX Version 6.1, TL06, SP5. this was released in 2011 week 15 and is, I believe, what operations is using. They are VERY conservative about updating OS patches in production and will only patch when faced with known problems. the complication is, I can't reproduce the problem in my lab, i think I disabled ipv6 entirely a long time ago.... the problem is occurring in our production operations overseas, which have a 12-14 hour time delta, when I've tried to work with them to match up the settings, the time and communications barriers have made it very difficult. once they get a workaround that gets past this problem, I tend to not hear from them again until there is a new problem. I'm suggesting the local admin try listen_addresses = '0.0.0.0' fix rather than listing 127.0.0.1,ip.of.machine, lets see if this works for us. If you can point me to a specific patch level (TL, SP) for that getaddrinfo problem, I can ask operations to test it on their staging environment. -- john r pierce N 37, W 122 santa cruz ca mid-left coast