Re: Connection Manager and I4GL

Keith Simmons wrote: > 2009/8/14 Jonathan Leffler <jonathan.leffler@gmail.com>: >> On Aug 13, 6:16 am, Keith Simmons <smile...@googlemail.com> wrote: >>> All >>> >>> Following 'recent' issues (this week and 15 months ago !!) management >>> have decided we need some fail-over mechanism in place for when the >>> database 'locks up'. >>> IDS HDR is the obvious candidate but I need to get my head around how >>> an application can seamlessly fail-over from primary to secondary. I'm >>> investigating Connection Manager for our ancillary applications that >>> use ODBC/JDBC/PHP etc., but the application that runs our busines is >>> written in I4GL running on 'green screens'. My understanding is that >>> Connection Manager cannot be used with I4GL, is this correct? If so >>> does any one know of an application that provides similar >>> functionality to enable auto-switching between databases, how do other >>> people get around this issue? >> >> I4GL 7.50 is built using CSDK 3.50 and should be able to use >> Connection Manager without problem. >> >> Older versions of I4GL use older versions of ESQL/C (or CSDK) and will >> not work with Connection Manager. >> >> -=JL=- >> >> _______________________________________________ >> Informix-list mailing list >> Informix-list@iiug.org >> http://www.iiug.org/mailman/listinfo/informix-list >> > Many thanks to all my correspondants. > I feel a test setup of two or three (virtual) servers coming up to > investigate all this before risking my live environment. A couple of > flies in the ointment is that the application is a third party > supplied package and that all the GL is currently run as interpreted > rather than compiled. We may need to change this !! As an alternative you could handle this through the IP addresses. Have the standby change its address to that of the main. When the previous main is repaired you could either allocate it to the standby address or switch the standby machine back to its original address. -- Ian Hotmail is for spammers. Real mail address is igoddard at nildram co uk

On 14/08/2009 12:25, Ian Goddard wrote: > As an alternative you could handle this through the IP addresses. Have > the standby change its address to that of the main. When the previous > main is repaired you could either allocate it to the standby address or > switch the standby machine back to its original address. > IMHO this is a 'Bad Idea' (tm) a) It will confuse the shit out of admin people b) You would have to wait for the arp caches to flush or do it manually If you are going to go down this route then at least have a floating IP address for the 'active' engine - don't change the primary IP of a given piece of tin. If you are on linux take a look at linux-ha - it would allow you to automate many of the tasks. -- Clive

Clive Eisen wrote: > On 14/08/2009 12:25, Ian Goddard wrote: > >> As an alternative you could handle this through the IP addresses. Have >> the standby change its address to that of the main. When the previous >> main is repaired you could either allocate it to the standby address or >> switch the standby machine back to its original address. >> > > IMHO this is a 'Bad Idea' (tm) > > a) It will confuse the shit out of admin people Admin people would have to put it into place. They're paid to look after stuff like this. > b) You would have to wait for the arp caches to flush or do it manually Agreed. It wouldn't be invisible. But it's easier than going round all the clients changing them to connect to a different host if Connection Manager isn't usable; especially so when some of the clients are a couple of hundred miles away. > > If you are going to go down this route then at least have a floating IP > address for the 'active' engine - don't change the primary IP of a given > piece of tin. > > If you are on linux take a look at linux-ha - it would allow you to > automate many of the tasks. HP-UX had an add-on something like this which is what prompted the thought. For all I know they might still have it. I think Keith said somewhere that he was on AIX, maybe IBM have their own solution. -- Ian Hotmail is for spammers. Real mail address is igoddard at nildram co uk

Hi all, the HP-UX High Availability solution is called MC/Service Guard - this product is still available and also supported on Linux! So, if you need a real High Availability solution you should go with MC/Service Guard (including transparent CPU mirroring, disk mirroring, IP address mirroring, automatic switch over, ...). Regards, Andreas Maier > -----Ursprüngliche Nachricht----- > Von: Ian Goddard [mailto:goddai01@hotmail.co.uk] > Gesendet: Freitag, 14. August 2009 14:04 > An: informix-list@iiug.org > Betreff: Re: Connection Manager and I4GL > > > Clive Eisen wrote: > > On 14/08/2009 12:25, Ian Goddard wrote: > > > >> As an alternative you could handle this through the IP addresses. > >> Have the standby change its address to that of the main. When the > >> previous main is repaired you could either allocate it to > the standby > >> address or switch the standby machine back to its original address. > >> > > > > IMHO this is a 'Bad Idea' (tm) > > > > a) It will confuse the shit out of admin people > > Admin people would have to put it into place. They're paid to look > after stuff like this. > > > b) You would have to wait for the arp caches to flush or do it > > manually > > Agreed. It wouldn't be invisible. But it's easier than > going round all > the clients changing them to connect to a different host if > Connection > Manager isn't usable; especially so when some of the clients are a > couple of hundred miles away. > > > > > If you are going to go down this route then at least have a > floating > > IP > > address for the 'active' engine - don't change the primary > IP of a given > > piece of tin. > > > > If you are on linux take a look at linux-ha - it would allow you to > > automate many of the tasks. > > HP-UX had an add-on something like this which is what prompted the > thought. For all I know they might still have it. I think > Keith said > somewhere that he was on AIX, maybe IBM have their own solution. > > -- > Ian > > Hotmail is for spammers. Real mail address is igoddard > at nildram co uk _______________________________________________ > Informix-list mailing list > Informix-list@iiug.org > http://www.iiug.org/mailman/listinfo/informix-list > >

2009/8/14 Ian Goddard <goddai01@hotmail.co.uk>: > Clive Eisen wrote: >> On 14/08/2009 12:25, Ian Goddard wrote: >> >>> As an alternative you could handle this through the IP addresses.  Have >>> the standby change its address to that of the main.  When the previous >>> main is repaired you could either allocate it to the standby address or >>> switch the standby machine back to its original address. >>> >> >> IMHO this is a 'Bad Idea' (tm) >> >> a) It will confuse the shit out of admin people > > Admin people would have to put it into place.  They're paid to look > after stuff like this. > >> b) You would have to wait for the arp caches to flush or do it manually > > Agreed.  It wouldn't be invisible.  But it's easier than going round all > the clients changing them to connect to a different host if Connection > Manager isn't usable; especially so when some of the clients are a > couple of hundred miles away. > >> >> If you are going to go down this route then at least have a floating IP >> address for the 'active' engine - don't change the primary IP of a given >> piece of tin. >> >> If you are on linux take a look at linux-ha - it would allow you to >> automate many of the tasks. > > HP-UX had an add-on something like this which is what prompted the > thought.  For all I know they might still have it.  I think Keith said > somewhere that he was on AIX, maybe IBM have their own solution. > > -- > Ian > > Hotmail is for spammers.  Real mail address is igoddard > at nildram co uk > _______________________________________________ > Informix-list mailing list > Informix-list@iiug.org > http://www.iiug.org/mailman/listinfo/informix-list > I am the admin, but as the DR machine is on a different physical site with separate sub-netting this is a non-starter. AIX !!! so linux/hp solutions are no good. There is not going to be a quick or easy resolution for this issue, need to do a bit more investigation/testing. Keith