Re: OPS Sequences: nocache == order ??
1) I think that the tpc numbers are done represented in Transactons Per Minute (TPM/C) and not Per second. So event with 50 tpm/c it means around 8000 tps. 2) Inserting 13000 rows with direct I/O doesn't mean you did 13000 transactions. It could be one transaction 3) I have seen the theoretical limit, but if I recall correctly it was the number of SCN numbers that was generated. That way they can calculate how long it takes before the SCN number will wrap (it is only 48 bits). That is way in the future. Anjo. On Saturday 07 September 2002 04:08, you wrote: One of our accelerator control system developers, an Oracle neophyte, claims to have achieved 13,000 tps writing to a RAID 5 array. I did set up the database, but most of the credit goes to him for exploring the OCI direct I/O options. I have no verified the rate, but I have no reason whatsoever to doubt him. This is on older four processor sun box. We've now traded in the lone a-1000 ,attached two T3's, and turned on archive logging. I had him retest and he said it was quicker than before . It's still RAID 5. If you are wondering why RAID 5, we have another little 659.9 Terabyte database and thousands of machines in compute farms to process the associated data. That project has first choice, and the rest of us make do with what's left. I too am curious where this theoretical limit of 16384 comes from. Theoretical as it no matter what hardware one chose this limit could not be surpassed? Ian MacGregor Stanford Linear Accelerator Center [EMAIL PROTECTED] -Original Message- Sent: Friday, September 06, 2002 4:38 PM To: Multiple recipients of list ORACLE-L On Wednesday 04 September 2002 09:53, Tim Gorman wrote: Thinking more about it last night... Since Oracle's theoretical limit is 16384 commits per second, I imagine that you could safely make the sequence recycle at (or 16384 or 9) and limit the number of digits contributed by the sequence to 4-5... Really? What have they done in the past to get those astronomical TPS numbers on some of their bencmarks? I'm pretty sure they were in excess of that number. IIRC, they were done on an nCube using OPS and about 400 CPUs. Jared -- Please see the official ORACLE-L FAQ: http://www.orafaq.com -- Author: Anjo Kolk INET: [EMAIL PROTECTED] Fat City Network Services-- (858) 538-5051 FAX: (858) 538-5051 San Diego, California-- Public Internet access / Mailing Lists To REMOVE yourself from this mailing list, send an E-Mail message to: [EMAIL PROTECTED] (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).
Re: Same SQL statement, Same Oracle, Different OS == Different E
I don't have an explanation, but: The CBO uses floating point instructions a lot (I think). May be there is a difference in rounding between the Intel chip and the Sparc chip that causes the cost to differ just enough so that the index isn't/is used. Anjo. On Saturday 07 September 2002 01:48, you wrote: Interesting. This might explain a similar problem I had a few years ago. Oracle support did not have a good answer for it. Jared On Friday 06 September 2002 13:43, Toepke, Kevin M wrote: I the RBO, the order the indexes were created in is important! I was able to show this to management on a project I was on. How? By doing a difinitive proof (follows) Import the table and data into an empty database. Create index A Create index B EXPLAIN PLAN shows query using index A. Drop table Import the table and data into an empty database Create index B Create index A EXPLAIN PLAN shows query using index B. Drop table Import the table and data into an empty database Create index A Create index B EXPLAIN PLAN shows query using index A. All other things being equal, the RBO will choose the index with the lower object_id! Proof took place in Oracle 8.0.5 on a Sun Solaris box. Kevin -Original Message- Sent: Friday, September 06, 2002 3:28 PM To: Multiple recipients of list ORACLE-L (see answer below) -Original Message- From: Sam Bootsma [ mailto:[EMAIL PROTECTED] mailto:[EMAIL PROTECTED] ] One of our developers is encountering a situation where Oracle 9.0.x explain plan chooses one index when on UNIX, and chooses a different index when running on Windows NT. I'd appreciate any insights or similar experiences. The following are the facts: 1. The explain plan is run against the same SELECT statement on both platforms 2. It has been confirmed that there are no statistics on either of the databases 3. Both databases are using RBO (not CBO) 4. The UNIX database has about 100 times as many rows (in this table) as the NT database 5. The SELECT statement that gives different explain plans on different platforms is: SELECT FN.*, FN.ROWID FROM UNITFUND FN WHERE FN.FU_CODE = :cFuCode AND FN.MKEY = :cMkey AND FN.CLNT = :sKey AND FN.PLANNO = :sKey AND FN.DATE_FROM = :dDate AND FN.SOURCE = :cSource AND FN.TSTATUS = 'O' ORDER BY FN.DATE_FROM, FN.TSECOND; 6. Between the following 2 indexes, Oracle 9.0x chooses (2) on Unix and (1) on Windows NT. 1) clnt, mkey, planno, fu_code, date_from, source, tracode, tsecond... 2) date_from, clnt, planno, mkey, fu_code Just a wild guess, but maybe the optimizer is just picking the first index it finds because it thinks both are equally good candidates. Were both indexes created in the same order on both databases? I.e. is Object_id (from dba_objects) for Index A smaller than object_id for Index B on the UNIX database, but the reverse is true on Windows? You say the databases are using RBO. How do you know? Remember that if you use some new features (from the manual: Partitioned tables and indexes Index-organized tables Reverse key indexes Function-based indexes SAMPLE clauses in a SELECT statement Parallel query and parallel DML Star transformations and star joins Extensible optimizer Query rewrite with materialized views Enterprise Manager progress meter Hash joins Bitmap indexes and bitmap join indexes Index skip scans ) the query optimizer will use CBO because new features are not supported by RBO. Finally, this section of the manual may help you guess what's happening: Oracle9i Database Performance Guide and Reference Part Number A87503-02 Chapter 8 Using the Rule-Based Optimizer ... Understanding Access Paths for the RBO Content-Type: text/html; charset=iso-8859-1; name=Attachment: 1 Content-Transfer-Encoding: 7bit Content-Description: -- Please see the official ORACLE-L FAQ: http://www.orafaq.com -- Author: Anjo Kolk INET: [EMAIL PROTECTED] Fat City Network Services-- (858) 538-5051 FAX: (858) 538-5051 San Diego, California-- Public Internet access / Mailing Lists To REMOVE yourself from this mailing list, send an E-Mail message to: [EMAIL PROTECTED] (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).
Re: OPS Sequences: nocache == order ??
Ok, It is saturday morning (brain is working at half power), I have looked at this function and at the original requirement and see a problem (may be 2). 1) The time of the multiple instances needs to be in sync with each other. If not, it could be that the 2nd instance has an earlier time and insert a record with a lower number after an insert of a higher number. So the real order is lost (that was a requirement). 2) Given the fact that the sequence numbers may be cached, even when the time is in sync, depending on the cached sequence numbers you could still end up with one instance inserting a number with a higher sequence number before the other instance with a lower sequence number in the same time (at seconds level). So if the requirements aren't so strict, why not drop the 'no order' and bump the cache ? Again, I may have missed something. Anjo. On Wednesday 04 September 2002 08:28, you wrote: Mladen, Is there any way to have developers/users access the sequence via a function, instead of accessing the sequence directly? If so, then perhaps you could modify the sequence to add the temporal component, while maintaining the use of a cached sequence for uniqueness? Such as: SQL create or replace function gen_seqq(in_seq in number) 2 return number 3 as 4 v_return_nbr number; 5 begin 6 select to_number(to_char(sysdate,'MMDDHH24MISS')||ltrim(to_char(in_seq,'00 00'))) 7 into v_return_nbr 8 from dual; 9 return v_return_nbr; 10* end gen_seqq; SQL / Function created. SQL create table x (y number); Table created. SQL create sequence xq; Sequence created. SQL insert into x values (gen_seqq(xq.nextval)); 1 row created. SQL Big and ugly numbers yes, but I think some folks get a strange thrill out of 20-digit numbers. It fits the requirement of being temporal (to the second, at least) and unique. You can throw in HSECS from V$TIMER if someone gets picky enough to want to go to the centi-second level as well. Yeah, and you can throw in USERENV('INSTANCEID') too, just for some real OPS/RAC-ness! Best of all, it fits the DBA-half of your brain by being fully cacheable and non-pinging... ...of course, you can embed the use of the SEQUENCE object inside the function; I left it on the outside in this example just to make it more flexible with regard to which sequence object it uses... If they don't like the idea of using a stored function to get the sequence number, then tell 'em that it's more ANSI standard that way and it's database independent. That gets 'em every time... Hope this helps... -Tim - Original Message - To: Multiple recipients of list ORACLE-L [EMAIL PROTECTED] Sent: Tuesday, September 03, 2002 5:54 PM Unfortunately, we have an application dependency and I was required to come up with a quick dirty fix. Thanks for your reply. On 2002.09.03 19:10 Anjo Kolk wrote: If you run OPS and specify order, it works like no cache. My question to you: Why cripple OPS and your business performance by having this requirement ? Spending a few bucks to get rid of this dependency will improve the performance, until you run in to the next problem ;-) Anjo. On Wednesday 04 September 2002 00:00, you wrote: I'm managing an OPS configuration (4x HP 9000/N, HP-UX 11/64 , RDBMS 8.1.7.1) and I'm having an application dependency on a temporal order of sequence numbers. With OPS that becomes a problem because each node caches a set of sequence numbers (20 by default). Oracle has an option, specifically for that situation, namely ORDER. My question is whether ORDER is the same thing as NOCACHE and whether it is possible to have a NOCACHE sequence which will return numbers in an incorrect order (larger number before the smaller one). Please, o OPS gods and godesses, help me out and I'll sacrifice you a beer when I see you. Mladen Gogala -- Please see the official ORACLE-L FAQ: http://www.orafaq.com -- Author: Anjo Kolk INET: [EMAIL PROTECTED] Fat City Network Services-- (858) 538-5051 FAX: (858) 538-5051 San Diego, California-- Public Internet access / Mailing Lists To REMOVE yourself from this mailing list, send an E-Mail message to: [EMAIL PROTECTED] (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 -- Please see the official ORACLE-L FAQ: http://www.orafaq.com -- Author: Mladen Gogala INET: [EMAIL PROTECTED] Fat City Network Services-- (858) 538-5051 FAX: (858)
Re: Oracle to publish pricing guide on Sept. 3
Yep, Open source, it is free at last ! On Saturday 07 September 2002 04:18, you wrote: ... and as September 6th dwindles into its waning hours, there is still no pricing guide. Jared On Wednesday 04 September 2002 13:59, [EMAIL PROTECTED] wrote: Just checked this, and it is supposed to be out now on Sept 6th. Jared DENNIS WILLIAMS [EMAIL PROTECTED] Sent by: [EMAIL PROTECTED] 08/29/2002 01:03 PM Please respond to ORACLE-L To: Multiple recipients of list ORACLE-L [EMAIL PROTECTED] cc: Subject:RE: Oracle to publish pricing guide on Sept. 3 Oracle has moved the release date from Aug. 28 to Sept. 3. All will be revealed . . . http://www.eweek.com/article2/0,3959,491399,00.asp -Original Message- Sent: Wednesday, August 28, 2002 2:02 PM To: '[EMAIL PROTECTED]' Has anyone seen this pricing guide yet? http://www.infoworld.com/articles/hn/xml/02/08/16/020816hnoraguide.xml?08 16 f ram -- Please see the official ORACLE-L FAQ: http://www.orafaq.com -- Author: Anjo Kolk INET: [EMAIL PROTECTED] Fat City Network Services-- (858) 538-5051 FAX: (858) 538-5051 San Diego, California-- Public Internet access / Mailing Lists To REMOVE yourself from this mailing list, send an E-Mail message to: [EMAIL PROTECTED] (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).
How to make web server to recognize another server location.
Title: How to make web server to recognize another server location. Hi, we have installed Forms 6i web server. In our applications we need to retrieve images stored in another server. The issue is that the web server does not recognize the other destinations on the network. When we copy the images to the web server, it works fine. How to make web server recognize the other servers on the network so that data can be accessed? rgds amar http://amzone.netfirms.com
RE: OPS Sequences: nocache == order ??
As I said I didn't verify his figures nor confirm he understands what constitutes a transaction. I'll endeavor to do so. The system collects data from monitors measuring the health of various test accelerator equipment. The telemetry is buffered before being inserted so that multiple readings could be pushed to the database as a single transaction which buoys Anjo's thought that 13,000 rows per second are being inserted not 13,000 tps. Ian MacGregor Stanford Linear Accelerator Center [EMAIL PROTECTED] -Original Message- Sent: Saturday, September 07, 2002 1:13 AM To: Multiple recipients of list ORACLE-L 1) I think that the tpc numbers are done represented in Transactons Per Minute (TPM/C) and not Per second. So event with 50 tpm/c it means around 8000 tps. 2) Inserting 13000 rows with direct I/O doesn't mean you did 13000 transactions. It could be one transaction 3) I have seen the theoretical limit, but if I recall correctly it was the number of SCN numbers that was generated. That way they can calculate how long it takes before the SCN number will wrap (it is only 48 bits). That is way in the future. Anjo. On Saturday 07 September 2002 04:08, you wrote: One of our accelerator control system developers, an Oracle neophyte, claims to have achieved 13,000 tps writing to a RAID 5 array. I did set up the database, but most of the credit goes to him for exploring the OCI direct I/O options. I have no verified the rate, but I have no reason whatsoever to doubt him. This is on older four processor sun box. We've now traded in the lone a-1000 ,attached two T3's, and turned on archive logging. I had him retest and he said it was quicker than before . It's still RAID 5. If you are wondering why RAID 5, we have another little 659.9 Terabyte database and thousands of machines in compute farms to process the associated data. That project has first choice, and the rest of us make do with what's left. I too am curious where this theoretical limit of 16384 comes from. Theoretical as it no matter what hardware one chose this limit could not be surpassed? Ian MacGregor Stanford Linear Accelerator Center [EMAIL PROTECTED] -Original Message- Sent: Friday, September 06, 2002 4:38 PM To: Multiple recipients of list ORACLE-L On Wednesday 04 September 2002 09:53, Tim Gorman wrote: Thinking more about it last night... Since Oracle's theoretical limit is 16384 commits per second, I imagine that you could safely make the sequence recycle at (or 16384 or 9) and limit the number of digits contributed by the sequence to 4-5... Really? What have they done in the past to get those astronomical TPS numbers on some of their bencmarks? I'm pretty sure they were in excess of that number. IIRC, they were done on an nCube using OPS and about 400 CPUs. Jared -- Please see the official ORACLE-L FAQ: http://www.orafaq.com -- Author: Anjo Kolk INET: [EMAIL PROTECTED] Fat City Network Services-- (858) 538-5051 FAX: (858) 538-5051 San Diego, California-- Public Internet access / Mailing Lists To REMOVE yourself from this mailing list, send an E-Mail message to: [EMAIL PROTECTED] (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). -- Please see the official ORACLE-L FAQ: http://www.orafaq.com -- Author: MacGregor, Ian A. INET: [EMAIL PROTECTED] Fat City Network Services-- (858) 538-5051 FAX: (858) 538-5051 San Diego, California-- Public Internet access / Mailing Lists To REMOVE yourself from this mailing list, send an E-Mail message to: [EMAIL PROTECTED] (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).
RE: Correlated subquery performance in 8i 9i
How about a little more info? I'm kind of reading between the lines here but the comments about correlated queries and things changing in 9i make me wonder if Gaja was getting at the ability to exploit hash/merge joins for correlated queries. For example: 1 select * 2 from code_master cm 3 where exists (select null 4from code_detail cd 5* where cm.code = cd.code) SQL / Execution Plan -- 0 SELECT STATEMENT Optimizer=CHOOSE (Cost=65 Card=5000 Bytes=65000) 10 FILTER 21 TABLE ACCESS (FULL) OF 'CODE_MASTER' (Cost=65 Card=5000 Bytes=65000) 31 INDEX (RANGE SCAN) OF 'CODE_DETAIL_PK' (UNIQUE) (Cost=3 Card=3 Bytes=12) For each row returned from code_master, a correlated query would be issued against code_detail (using an index). Now, if I put this in an in-line view, I can get a hash-join approach, which is desired in this case since I am getting every row from each table -- I would *not* want a correlated approach in this case, I would prefer a hash approach: SQL select * 2 from code_master cm, 3 (select code 4 from code_detail) cd 5 where cm.code = cd.code 6 / Execution Plan -- 0 SELECT STATEMENT Optimizer=CHOOSE (Cost=1106 Card=299600 Bytes=5093200) 10 HASH JOIN (Cost=1106 Card=299600 Bytes=5093200) 21 TABLE ACCESS (FULL) OF 'CODE_MASTER' (Cost=340 Card=10 Bytes=130) 31 INDEX (FAST FULL SCAN) OF 'CODE_DETAIL_PK' (UNIQUE) (Cost=131 Card=299600 Bytes=1198400) So, by going to the in-line view, I get better performance, but this doesn't necessarily mean it is best for all queries, just this query. Now, about 9i and how things change. I wonder if Gaja was referring to always_semi_join becoming an undocumented parameter _always_semi_join and the CBO deciding whether to un-correlate a correlated query and use a hash approach. Note that this is also available in 8i if you set always_semi_join = hash, or use the semi join hints (hash_sj / merge_sj), and the conditions for a semi join are met (and if not, the in-line view approach is a good workaround): SQL alter session set always_semi_join = hash; Session altered. SQL select * 2 from code_master cm 3 where exists (select null 4from code_detail cd 5where cm.code = cd.code) 6 / Execution Plan -- 0 SELECT STATEMENT Optimizer=CHOOSE (Cost=831 Card=99900 Bytes=1698300) 10 HASH JOIN (SEMI) (Cost=831 Card=99900 Bytes=1698300) 21 TABLE ACCESS (FULL) OF 'CODE_MASTER' (Cost=65 Card=10 Bytes=130) 31 INDEX (FAST FULL SCAN) OF 'CODE_DETAIL_PK' (UNIQUE) (Cost=131 Card=299600 Bytes=1198400) Note that a correlated approach for the exists is no longer used. A hash approach, similar to that when using an in-line technique, is available. Now in 9i, the CBO will choose to stay correlated, or, to un-correlate and use a hash/merge approach, if you do not set the _always_semi_join parameter. SQL connect scott/tiger@or91 Connected. SQL set autotrace trace explain SQL select * 2 from code_master cm 3 where exists (select null 4from code_detail cd 5where cm.code = cd.code) 6 / Execution Plan -- 0 SELECT STATEMENT Optimizer=CHOOSE (Cost=770 Card=99900 Bytes=1498500) 10 HASH JOIN (SEMI) (Cost=770 Card=99900 Bytes=1498500) 21 TABLE ACCESS (FULL) OF 'CODE_MASTER' (Cost=77 Card=10 Bytes=110) 31 INDEX (FAST FULL SCAN) OF 'CD_CODE_IDX' (NON-UNIQUE) (Cost=208 Card=299600 Bytes=1198400) Note that the CBO chose a hash approach for the above example. Now, if I do something highly selective on code_master (foo_date = trunc(sysdate) - 200), where I *would* want a correlated approach, note that the CBO chooses the correlated approach as opposed to the hash approach above: 1 select * 2 from code_master cm 3 where exists (select null 4from code_detail cd 5where cm.code = cd.code) 6* and foo_date = trunc(sysdate) - 200 SQL / Execution Plan -- 0 SELECT STATEMENT Optimizer=CHOOSE (Cost=6 Card=1 Bytes=15) 10 NESTED LOOPS (SEMI) (Cost=6 Card=1 Bytes=15) 21 TABLE ACCESS (BY INDEX ROWID) OF 'CODE_MASTER' (Cost=4 Card=1 Bytes=11) 32 INDEX (RANGE SCAN) OF 'CM_FD_IDX' (NON-UNIQUE) (Cost=3 Card=1) 41 INDEX (RANGE SCAN) OF 'CD_CODE_IDX' (NON-UNIQUE) (Cost=2 Card=299600 Bytes=1198400) Anyway, Gaja, are you out there? I'm interested in what this person was referring to. Reading between the lines of what the person posted, I am guessing that he was referring to the ability to get a hash join, when
No space left on device - but I have lots left.
Hello guys, Could use some help here. 1) System specs. Solaris 8 512MB Ram 1GB Swap Oracle 9i Rel2 2) Relevant init.ora parameters: *.LARGE_POOL_SIZE=1048576 *.SGA_MAX_SIZE=33554432 *.SHARED_POOL_SIZE=16777216 *.SORT_AREA_SIZE=2097152 *.compatible='9.2.0.1.0' *.db_cache_size=2097152 *.log_buffer=1048576 3) Starting up database, I get the following error: SQL startup nomount ORA-27302: failure occurred at: skgpwreset1 ORA-27303: additional information: invalid shared ctx ORA-27146: post/wait initialization failed ORA-27300: OS system dependent operation:semget failed with status: 28 ORA-27301: OS failure message: No space left on device ORA-27302: failure occurred at: sskgpsemsper I have a lot of disk space left - 2.5GB. I have half of my RAM and the whole swap free when this error occurs. I appreciate any help and tips. Thank you. ltiu -- Please see the official ORACLE-L FAQ: http://www.orafaq.com -- Author: ltiu INET: [EMAIL PROTECTED] Fat City Network Services-- (858) 538-5051 FAX: (858) 538-5051 San Diego, California-- Public Internet access / Mailing Lists To REMOVE yourself from this mailing list, send an E-Mail message to: [EMAIL PROTECTED] (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).
Re: No space left on device - but I have lots left.
I love emails like this. Do you read the error message at all or you simply email all your findings to the list assuming that the list will automatically be interested in everything coming from you? Don't you think that starting your email with something like sorry guys and goddesses, I'm a beginner in dire need would be in order? Now to your message. You need more semaphores, not disk space. Get a copy of Adrian Cockroft's performance tuning book and it will tell you exactly how to modify your /etc/system. Also, you may try looking for the words like semaphore at docs.sun.com. I know it's unusual thing to do and much harder then to email this list but you might want to try it nevertheless. On 2002.09.07 18:43 ltiu wrote: Hello guys, Could use some help here. 1) System specs. Solaris 8 512MB Ram 1GB Swap Oracle 9i Rel2 2) Relevant init.ora parameters: *.LARGE_POOL_SIZE=1048576 *.SGA_MAX_SIZE=33554432 *.SHARED_POOL_SIZE=16777216 *.SORT_AREA_SIZE=2097152 *.compatible='9.2.0.1.0' *.db_cache_size=2097152 *.log_buffer=1048576 3) Starting up database, I get the following error: SQL startup nomount ORA-27302: failure occurred at: skgpwreset1 ORA-27303: additional information: invalid shared ctx ORA-27146: post/wait initialization failed ORA-27300: OS system dependent operation:semget failed with status: 28 ORA-27301: OS failure message: No space left on device ORA-27302: failure occurred at: sskgpsemsper I have a lot of disk space left - 2.5GB. I have half of my RAM and the whole swap free when this error occurs. I appreciate any help and tips. Thank you. ltiu -- Please see the official ORACLE-L FAQ: http://www.orafaq.com -- Author: ltiu INET: [EMAIL PROTECTED] Fat City Network Services-- (858) 538-5051 FAX: (858) 538-5051 San Diego, California-- Public Internet access / Mailing Lists To REMOVE yourself from this mailing list, send an E-Mail message to: [EMAIL PROTECTED] (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 -- Please see the official ORACLE-L FAQ: http://www.orafaq.com -- Author: Mladen Gogala INET: [EMAIL PROTECTED] Fat City Network Services-- (858) 538-5051 FAX: (858) 538-5051 San Diego, California-- Public Internet access / Mailing Lists To REMOVE yourself from this mailing list, send an E-Mail message to: [EMAIL PROTECTED] (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).
Re: No space left on device - but I have lots left.
Thank you. Yes. You are right. The error messages are very informative don't you think? ltiu On Saturday 07 September 2002 16:23, Mladen Gogala wrote: I love emails like this. Do you read the error message at all or you simply email all your findings to the list assuming that the list will automatically be interested in everything coming from you? Don't you think that starting your email with something like sorry guys and goddesses, I'm a beginner in dire need would be in order? Now to your message. You need more semaphores, not disk space. Get a copy of Adrian Cockroft's performance tuning book and it will tell you exactly how to modify your /etc/system. Also, you may try looking for the words like semaphore at docs.sun.com. I know it's unusual thing to do and much harder then to email this list but you might want to try it nevertheless. On 2002.09.07 18:43 ltiu wrote: Hello guys, Could use some help here. 1) System specs. Solaris 8 512MB Ram 1GB Swap Oracle 9i Rel2 2) Relevant init.ora parameters: *.LARGE_POOL_SIZE=1048576 *.SGA_MAX_SIZE=33554432 *.SHARED_POOL_SIZE=16777216 *.SORT_AREA_SIZE=2097152 *.compatible='9.2.0.1.0' *.db_cache_size=2097152 *.log_buffer=1048576 3) Starting up database, I get the following error: SQL startup nomount ORA-27302: failure occurred at: skgpwreset1 ORA-27303: additional information: invalid shared ctx ORA-27146: post/wait initialization failed ORA-27300: OS system dependent operation:semget failed with status: 28 ORA-27301: OS failure message: No space left on device ORA-27302: failure occurred at: sskgpsemsper I have a lot of disk space left - 2.5GB. I have half of my RAM and the whole swap free when this error occurs. I appreciate any help and tips. Thank you. ltiu -- Please see the official ORACLE-L FAQ: http://www.orafaq.com -- Author: ltiu INET: [EMAIL PROTECTED] Fat City Network Services-- (858) 538-5051 FAX: (858) 538-5051 San Diego, California-- Public Internet access / Mailing Lists To REMOVE yourself from this mailing list, send an E-Mail message to: [EMAIL PROTECTED] (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). -- Please see the official ORACLE-L FAQ: http://www.orafaq.com -- Author: ltiu INET: [EMAIL PROTECTED] Fat City Network Services-- (858) 538-5051 FAX: (858) 538-5051 San Diego, California-- Public Internet access / Mailing Lists To REMOVE yourself from this mailing list, send an E-Mail message to: [EMAIL PROTECTED] (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).
Re: No space left on device - but I have lots left.
-- ltiu [EMAIL PROTECTED] The error messages are very informative don't you think? Error messages in most products stink. If you don't learn to outsmart the developers and find what's wrong you will never be able to manage databases or operating systems. -- Steven Lembark 2930 W. Palmer Workhorse Computing Chicago, IL 60647 +1 800 762 1582 -- Please see the official ORACLE-L FAQ: http://www.orafaq.com -- Author: Steven Lembark INET: [EMAIL PROTECTED] Fat City Network Services-- (858) 538-5051 FAX: (858) 538-5051 San Diego, California-- Public Internet access / Mailing Lists To REMOVE yourself from this mailing list, send an E-Mail message to: [EMAIL PROTECTED] (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).
Re: No space left on device - but I have lots left.
Yes. Thank you for your advice. I do have to keep this in mind everyday. ltiu On Saturday 07 September 2002 17:28, Steven Lembark wrote: -- ltiu [EMAIL PROTECTED] The error messages are very informative don't you think? Error messages in most products stink. If you don't learn to outsmart the developers and find what's wrong you will never be able to manage databases or operating systems. -- Please see the official ORACLE-L FAQ: http://www.orafaq.com -- Author: ltiu INET: [EMAIL PROTECTED] Fat City Network Services-- (858) 538-5051 FAX: (858) 538-5051 San Diego, California-- Public Internet access / Mailing Lists To REMOVE yourself from this mailing list, send an E-Mail message to: [EMAIL PROTECTED] (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).