Re: Raw or Buffered device?

r.a.reissaus@risdi.com (Robert A. Reissaus) writes: > Paul Watson <paul@oninit.com> wrote in message news:<3F634E49.93BE0D28@oninit.com>... >> [cutting] >> > IT IS PERFECTLY SAFE to just bring the engine down, >> > > remove the symbolic link to /dev/dbroot and replace it with an identically >> > > named symbolic link to /dev/rdbroot! >> > >> > This how ever I do not agree with. Changing the symbolic link will >> > result in error : >> > >> > Cannot open chunk '/opt/DBA/Informix/Links/730/risdi08/primary/root'. >> > errno = 19 >> > oninit: Fatal error in shared memory initialization >> > >> > because the (old) link itself is stored in the instances systam >> > tables. So it would not be sufficient to 'simply' change the link. >> > Robert A. Reissaus >> >> You can point the link to anywhere you want, the engine has no idea >> where the link points to, only the link name > > > He people, I didn't INVENT the error. It occurs! Again and again and > again! On SCO, on HPUX, on AIX and on Linux. So stop the 'NONSENSE' > remarks unless you can really prove I am wrong, then present the > argument and after i have verified you remarks are true and I am wrong > (which quite frankly would surprise me enormously) i will admit i am > wrong. I think you will have to explain just HOW you did the relinking. I'm sure you got this error, but it must have been because of some error on your part. I have done this several times. No problems. -- 'yvind Gjerstad Systems dept Tollpost-Globe AS N-6301 'ndalsnes/Norway E-mail: ogj@tollpost.no Phone: +47 7122 6663 Fax: +47 7122 6694

This was done by me or on my behalf: )HPUX (L2000) with a hybrid of IBM/HP storage LVM managers using IFX 7.3xFC. )SCO OpenServer 5.06 with IFX 7.3x (PC with onboard raid and IDE) )Aix 5.(something) with IFC 7.3x )-by me very recently; namely when testing the error; on a test database in use since 1999- on a Linux box (SuSE 8.2 Pro) using raw devices. Funny thing was when changing the raw device I got another error then when linking directly against the block device. There were no errors on my part that weren't checked. After rebuilding/(re)designing not so small 24x7 production/warehouse databases (anywhere between 10 and 650GB) for allmost 10 years now that would be inexcusable. What I find funny is that all this has been said without *anyone*; including myself asking about the configuration (hardware/software/ifx version). No people; one of the first things I make absolutely sure to be correct is the link and the paths. (Now I am probably going to make myself very popular, but I am used to that) My point here is that it is never *SIMPLE* to change something, regardless of what *anyone* might say. And since we are in a newsgroup that many -newbies; more advanced; not well versed in english or whatever; frequently visit anyone replying (some people really should know better; and I am just a tad disappointed) *must* take that into account. Even after having given the same answers over and over again in the past years (which we no doubt all have done). This is typically one of the things that *usually should* go without a problem, but *sometimes* - if only because of a possible (undetectable) hardware (combination) failure- they go horribly wrong. So I think I (at least from my part) will consider this little item closed. Regards, till the next 'battle'. Keep it lively, mates! Robert A. Reissaus RISDI (tm)(R) Ibm/Informix Consultants Amsterdam, the Netherlands r.a.reissaus@risdi.com