Hi Neil,
I looked at the code as an Informix developer. I think so far it looks good. :)
But I have a few questions/comments:
- there seem to be 2 required parameters that are not quite clear to me, what
they are for and what values they should have: LH_RS and HAS_RS (in file
ids_config).
Is it possible to put some comments into this file to better explain the
purpose and possible/expected values for these two parameters?
- function check_start() (in file functions) seems to look for processes only
(using pgrep). This may not give the right "picture" if there is more than
one instance running on a machine.
[ Running more than one instances of IDS, even different versions of IDS,
may not be a very common thing, but is possible and perfectly legal. ]
In a future version if HA-Informix it may be worth thinking about this.
- there is a similar "problem" with identifying SHM segments (ipcs ...).
For one, there can be other SHM segments on a machine (with whatever
user/group owners), independent of IDS and user/group "informix".
But if there is more than one IDS instance running on a machine, then surely
there are more SHM segments on the machine - and only some of them will
belong to a particular IDS instance. Thus just doing "ipcs ..." will not
necessarily find out the state of a particular IDS instance correctly.
If you are interested, I can provide an algorithm to "reverse-engineer" the
SHM-key into the IDS SERVERNUM (as specified in the $ONCONFIG file). This
SERVERNUM is unambiguous for IDS instances (provided there is no Informix
XPS, for which the algorithm is a bit different).
- removal of SHM segments (in function cleanup_ipc()) is not always that easy.
As long as a process is still attached to an SHM segment, ipcrm will not
really
remove the SHM segment. Instead the attached process has to be killed first.
[ With IDS it is possible to connect a local client, e.g. the backup tools,
via specific SHM segments to the IDS instance (instead of using TCP/IP
connections). Even though the server may not be running anymore (
oninit processes died) and most SHM segments are gone, such a
communication SHM segment will not be removable as long as the local
client is attached to it. ]
With above algorithm to identify SHM segments of a specific IDS instance,
it would be possible to detect such a situation correctly and thus return
an error only when necessary (but not when un-related SHM segments are
still around).
Also, in the case of more than one instances on a machine, other instances'
SHM segments would probably get removed - effectively killing those instances
in the process.
That's it from my side.
TIA,
Martin.
--
Martin Fuerderer
IBM Informix Development Munich, Germany
Information Management
IBM Deutschland GmbH
Chairman of the Supervisory Board: Hans Ulrich M?rki
Board of Management: Martin Jetter (Chairman), Rudolf Bauer, Christian
Diedrich, Christoph Grandpierre, Matthias Hartmann, Thomas Fell, Michael Diemer
Corporate Seat: Stuttgart, Germany; Reg.-Gericht: Amtsgericht Stuttgart,
HRB-Nr.: 14 562 WEEE-Reg.-Nr. DE 99369940
--
This message posted from opensolaris.org
