Actually, 5 blocks wasn't completely hardwired, there was an undocumented
parameter ("_walk_insert_threshold" or something like that. My notes from
Scott Gosset's course are largely unreadable. What has hapened to my
handwriting? ) which was utilized to define the number of blocks that will be
added to the free list. If I remember corectly, there was a serious bug with
DMT, which was precisely about the FREELIST mechanism which would prevent
freed blocks from being reused. That was mentioned by somebody else here. I
believe that the workaround was to set _walk_nsert_threshold to 7. I
distinctly remember seeing K. Gopalakrishnan's name mentioned in connection
with that, and this is the same symptom that was described by somebody else on
this list. As I am utilizing LMT's and segment space auto management, I'm
mostly oblivious to DMT woes. K. Gopalakrishnan (I hope my spelling is
correct) my shed some more light onto this affair.
On 2003.10.01 00:59, Wolfgang Breitling wrote:
> I can't recall right now where I found out about the 3 blocks required for
> automatic space management. Could have been an error message when I tried to
> create a table with a 2 block extent in an ASSM tablespace, or a
> presentation at IOUG, or perhaps even on this list.
> The 5 block rule is the documented allocation rule for DMT where Oracle
> rounds requests for segments greater than 5 blocks to the next multiple of 5
> blocks (unless it finds a free segment of exactly the right size or ...).
> Which is why you couldn't implement a uniform extent size policy in DMT with
> extent sizes of exact powers of 2 (64, 128, 256, 1024, ...) to make use of
> the full IO bandwith of the OS (which is generally a power of 2) for full
> scans since they all were not multiples of 5 blocks. Not until the minimum
> extent size option came in Oracle 8 (not to be confused with minextents).
> But then LMTs came in Oracle 8i and retired the entire DMT allocation
> scheme.
>
> At 07:49 PM 9/30/2003 -0800, you wrote:
>
>> I repeated your test, with the same result. You, of course, are right.
>> Interesting, that means that oracle gave up on that "5 blocks rule".
>> Where did you come accross the fact that automatic space management
>> requires 3 blocks? That is, I suppose, for freeelists & freelist groups?
>> I must confess that I assumed that the old 5 blocks rule still holds true,
>> so I didn't test further. Also, I was testing the problem that I had with
>> autoallocate and automatic segment management, which turned out to be
>> a SCSI controller problem. Basically, when I created the tablespace on EIDE
>> device, it worked as advertised, but when I attempted to do that on a SCSI
>> disk, it failed. To dispell all doubts, SCSI controller died in 2 days,
>> causing, of course a system and the database crash. May it rest in peace,
>> in the place SCSI controllers go when they burn out.
>>
>> --
>> Mladen Gogala
>> Oracle DBA
>> --
>> Please see the official ORACLE-L FAQ: http://www.orafaq.net
>> --
>> Author: Mladen Gogala
>> INET: mgogala_at_adelphia.net
>
> Wolfgang Breitling
> Oracle7, 8, 8i, 9i OCP DBA
> Centrex Consulting Corporation
> http://www.centrexcc.com
>--
> Please see the official ORACLE-L FAQ: http://www.orafaq.net
>--
> Author: Wolfgang Breitling
> INET: breitliw_at_centrexcc.com
>
> Fat City Network Services -- 858-538-5051 http://www.fatcity.com
> San Diego, California -- Mailing list and web hosting services
> ---------------------------------------------------------------------
> To REMOVE yourself from this mailing list, send an E-Mail message
> to: ListGuru_at_fatcity.com (note EXACT spelling of 'ListGuru') and in
> the message BODY, include a line containing: UNSUB ORACLE-L
> (or the name of mailing list you want to be removed from). You may
> also send the HELP command for other information (like subscribing).
--
Mladen Gogala
Oracle DBA
--
Please see the official ORACLE-L FAQ: http://www.orafaq.net
--
Author: Mladen Gogala
INET: mgogala_at_adelphia.net
Fat City Network Services -- 858-538-5051 http://www.fatcity.com
San Diego, California -- Mailing list and web hosting services
---------------------------------------------------------------------
To REMOVE yourself from this mailing list, send an E-Mail message
to: ListGuru_at_fatcity.com (note EXACT spelling of 'ListGuru') and in
the message BODY, include a line containing: UNSUB ORACLE-L
(or the name of mailing list you want to be removed from). You may
also send the HELP command for other information (like subscribing).
Received on Wed Oct 01 2003 - 07:24:47 CDT
