startup fails after migration of Oracle DB 8i to 9i on Alpha 7.3-2

[Ian Miller.]

AFAIK The Account field is normally only used for accounting and reports and does not have much effect otherwise.

[Jan van den Ende]

Ian,

but,
the first time you enter a user in a group, specifying a name for the account field that does not yet translate into an identifier, that name is entered with the special identifier value of [,*] (and /attrib=resource, so it can take ownership etc)
If you thereafter add a username with the same name, then an identifier for that username is NOT created.
And now you are in for all kinds of "interesting" behavior for functionalities that depend on that identifier...

Proost.

Have one on me.

jpe

[Ian Miller.]

Jan, Do you mean if you add a user with a name equal to a name used as an account name? I can see this will cause problems.

I think AUTHORIZE trying to be helpful by adding the identifier with value [group,1777777].

You can modify users in a particular UIC group to have different account names if it suits your environment.

[Jan van den Ende]

Ian, yes.

Look back, THAT was what the OP planned, and what I was warning against.

Proost.

Have one on me.

jpe

[Guinaudeau]

Ian (Miller):

look at my previous remark about the confusion around ORAOMVN user and account name, as long as the account name was ORAOMVN in our user ORAOMV920.

Jan should be correct, it sounds like Oracle code CAN be confused. I think it has been confused. I will anyway pay attention in the future to these issues, with Oracle and with other apps we have.

Jan (the other one):

may be you quote here the link for this issue account/owner/user names in the documentation 7.3-2, thanks in advance. it will be helpfull to me and many, pretty sure.

i look at your suggestion(s). i have now compared our SYSUAF with the one of our colleages of the other department, here so called "they". attached extract of their SYSUAF, one VMS user ORACLE9 is used for both installation and DBA tasks.

they have only one DB on their test systems, and they installed directly 9i, they did not migrate from 8i.

they also have a Galaxy on DS25, we have a single VMS instance (test machine = DS10, customer machine = ES45). this should not be the issue, but :

=> did someone install oracle9i on DS10 machines ? could he confirm it's OK with our platform EV6 / SRM console V6.7-1 ?

they have actually used another setting than we had, and their settings match your advices in some sense.

their config :

1) one user only ORACLE9 is defined, used both to install the product
and to administrate the DB (incl start and stop instance)

2) this user has been defined without owner and account name, both strings are blank

=> i understand now that this UIC is used to create the files. i have been confused with the "owner" field of the SYSUAF configuration.

surprisingly, no SETPRV privilege in their user definition : this corresponds with Oracle9i documentation, but we got the explicit AUDIT privilege failure for it, so i dont understand what happened on our installation ...

3) they did not grant ORA_SGA to the ORACLE9 user : i overlooked that recommendation in Oracle9i installation guide (dec 2002, page 55) since we used to have this id granted to the installer+DBA account in Oracle8i

=> silly error of mine in our SYSUAF !!! will remove

4) they have defined but not granted ORA_ID ; i understand this could be forgotten, is very specific to their installation. i didn't find any trace of this ORA_ID ID under Oracle documentation and
Metalink, also.

5) they have defined and granted ORA_AGENT_ID but this should not be
significant / could be forgotten because the OEMAGENT does not work
properly under VMS.

i will compare further a GET_ENVIRONMENT (if you know that script) output of their machine for every SYSGEN parameters and other system informations etc ...

jan, since we had until now a group for different Oracle DBs, following your advice, i will define a new group for the Oracle9i user, will either not define explicitly owner and account (alike their example of
ORACLE9). or will define new owner and account names not used until now and different from the newly defined user name (alike your suggestion)

just we are waiting for better exchange of information and support between the two installations in our company (same company, but not same network, some stuff N/A just now now). should continue afterwards and update this thread.

thanks to you and every helpers until now. i am ever very pleased to attend this forum, never loose my time.

louis

nota -> jan : found a translator for your "have one on me" and might have the corresponding "HW" if you were close to Essen/Ruhr in Germany or would attend the next bootcamp.

[Wim Van den Wyngaert]

Just to be sure : if you put the account back to the old value, is the problem coming back ?

And fyi : numeric user names are very badly diggested because they are valid but no identifier is created (numeric identifiers names are not supported).

Wim

[Jan van den Ende]

Louis,

for your information: my LAST involvement with Oracle was V6.1-something, on Vax, way back in the times when Oracle was still a native VMS development.

All my remarks were in essence generic VMS remarks, not limited to Oracle per se, and I was not even sure _IF_ they apply in your specific case.
But, they ARE quite generic, and your symptoms seemed to fit, that is why I reacted.

The principle of a new UIC group for each new version of Oracle, with all installation and DBA activity by members of that group, used to be, and probably still is, the thing to do to avoid many problems that might otherwise arise.

Furthe specific Oracle details are probably better dealt with by current ORA users, which are not too rare in this forum.

About Bootcamp: Yes, I _DO_ intend to be there. When in Nashua, I am a regular at the hotel bar, ( just like Ian ), am if you locate me, and make yourself known, we will definitely share a Samuel Adams.

Proost.

Have one on me.

jpe

[Guinaudeau]

wim :

>>> if you put the account back to the old value, is the problem
coming back ?

cannot test this now now. will try this

>>> numeric user names are very badly diggested because they are valid
but no identifier is created (numeric identifiers names are not
supported).

do you mean with numeric user name alpha-num too alike "ORACLE9" ? what a surprise !

louis

[Jan van den Ende]

Louis,


do you mean with numeric user name alpha-num too alike "ORACLE9" ? what a surprise !


oh no, THOSE are perfectly acceptable.

But usernames like
Username: 12345
ARE valid usernames as well, but THOSE can NOT be associated with an identifier.

hth,

Proost.

Have one on me.

jpe

[Guinaudeau]

we first tried to migrate the existing DB.

we will try to run DBCA with the installed product, to compare the results. if new DB can be created, that is at least something and mean a correct product installation and account but something wrong with the migrated DB ! we would prefer migrate to
export/import for that large DB, for sure. (and we red one should forget DBCA tool itself ... sounds incorrect or misunderstood)

*************

ian + jan :

to summary, if i well understand you and the documentation around 732final\AA-PV5MH-TK\AA-PV5MH-TK.HTML :

please confirm or correct following strategy. My goal is to have a clear guideline (= 100% safe) to add a user account for a new Oracle release :

1) check whether the new user name is not already entered as an ID in SYSUAF

2) add the new user with a new group UIC : either w/o account name (alike the ORACLE9 example) or with account=user

=> this account name should not be confusing for VMS, because no previous user (= no existing ID) has been configured in SYSUAF with same name

=> this will add one unique ID, type alphanumeric, both for account name and user name

=> or did i misunderstood your explanation and should really add a new user name X with a new group UIC and a new account name Y ?

=> what about owner ? should i prefer to let owner undefined or to
have it = user name

3) continue with quota, password, privileges as requested by Oracle ; probably add SETPRV (although not documented by Oracle) : sounds to me the minimal list of privileges (auth+default) would be :

CMKRNL GROUP GRPNAM IMPERSONATE LOG_IO NETMBX
PFNMAP PRMCEB PRMGBL PRMMBX SETPRV SHARE
SYSGBL SYSLCK SYSNAM SYSPRV TMPMBX WORLD

4) in Oracle9i, at least, def and dir should be set such a way that it is a new path and new LOGIN.COM which does not run old ORAUSER env.

i dont care about ACL at that time : should i immediately care with them, since we intend eventually to have the Oracle tree to be read accessed as a Advanced Server share (prefer to see those UNIX-like paths and file names from PC, or ?) and we may add for this purpose an ACL ?