Fully compliant GIS
Mysql AB, What is the roadmap for fully compliant GIS? The OpenGIS specification defines the following functions, which MySQL does not yet implement. They should appear in future releases. When implemented, they will provide full support for spatial analysis, not just MBR-based support. Thanks, Trevor -- MySQL General Mailing List For list archives: http://lists.mysql.com/mysql To unsubscribe:http://lists.mysql.com/[EMAIL PROTECTED]
Re: MySQL and NPTL
Steve, I have had this same problem exept the entire server does not hang, only one thread which experiences a signficiat slowdown. This happens a couple times a week for our 4 servers. I believe this is a thread scheduling issue. The mysql support team suggests against using NPTL. However, I use mysqld-max on redhat linux 9.0 and have found a 10-20% performance difference at HIGH (hundreds of connections) load levels. The support team claims the difference in performance between linux threads and NPTL is never significant. Incidentally I have experinced this same thread starvation issue on redhat as 3.0 amd64 while using Linux threads. Theory is that glic 2.3 is to blame ( redhat 9.0 - glibc-2.3.2-11.9 , redhat as 3.0 - glibc-2.3.2-95.6 ). So the version of glibc looks to be the culprit. As I understand it Mysql is working with Redhat to determine the issues with glibc 2.3. I haven't tried the export line. Trevor Steve Meyers wrote: Has anyone else experienced this bug? http://bugs.mysql.com/bug.php?id=868 We've been seeing this problem on several of our servers (see the last comment to the bug). MySQL just hangs occasionally, it happens about 3-4 times per month. We have 13 database servers, so that unfortunately increases our odds. I just wondered how many other people have seen it, and if the suggested export LD_ASSUME_KERNEL=2.2.5; mysqld_safe has worked for anyone. In order to turn that on, I will need to take our site down completely, which is (of course) not desirable. Thanks! Steve Meyers -- MySQL General Mailing List For list archives: http://lists.mysql.com/mysql To unsubscribe:http://lists.mysql.com/[EMAIL PROTECTED]
Change ft_min_word_len require fulltext index rebuild
Mysqlians, I just discovered that a server has ft_min_word_len=4 when I want ft_min_word_len=3 . Do I have to rebuild all the fulltext indicies so that searches on three-letter words will work correctly? Trevor -- MySQL General Mailing List For list archives: http://lists.mysql.com/mysql To unsubscribe:http://lists.mysql.com/[EMAIL PROTECTED]
Re: avoiding Locked threads
Jon, an update statement is supposed to return the # of rows updated, so the client must wait for a response, which means the client will wait if you low_priority your updates. But if they are insert delayed then the client gets a return immediately. This behavior limits the usefullness of low priority with updates. I did ask on the list if there were any plans for a delayed update, but I was told there aren't. Trevor Jon Drukman wrote: Dathan Vance Pattishall wrote: log-bin=/var/opt/mysql/db2-binlog skip-innodb log-error=/var/opt/mysql/db2-errlog This is on a separate drive? yes, the database is the only thing on the high speed RAID. everything else is on the other drive (also a RAID but only RAID0 with 2 drives). any ideas appreciated! Try setting low-priority-updates and delay-key-write=ALL i haven't tried this yet, but one of the other developers has objected that doing this will kill performance for people posting messages because their clients will hang waiting for the selects to finish. is this true? Your running into a concurrency issue, the only other quick fix is to use innodb, but your blobs will kill you in disk space. h we've got approx 60G free on the RAID so this may not be such a big problem. -jsd- -- MySQL General Mailing List For list archives: http://lists.mysql.com/mysql To unsubscribe:http://lists.mysql.com/[EMAIL PROTECTED]
Maintaining fulltext
Mysqlians, 1. How can I tell how many words a fulltext index has indexed?(for purposes of estimating performance) 2. How can I tell how much index file space a full text is responsible for? ( for regular indicies I use (key_length +4)/.067 ) but my fulltext's are all on text's. Thanks, Trevor p.s. See you at the conference. -- MySQL General Mailing List For list archives: http://lists.mysql.com/mysql To unsubscribe:http://lists.mysql.com/[EMAIL PROTECTED]
Low_Priority Updates
Mysqlians, Does a low_priority behave like a DELAYED for insert and return to the caller before the update has actually been executed? If not are their plans for a DELAYED option with update? Regards, Trevor -- MySQL General Mailing List For list archives: http://lists.mysql.com/mysql To unsubscribe:http://lists.mysql.com/[EMAIL PROTECTED]
Read bursts on datadir
Mysqlians, Greetings We have been experiencing a performance problem that relates to either mysql's buffers of the ext3 file cache. Our mysql servers are experiencing load spikes due to massive disk reads on datadir( 3meg sec). Datadir is a seprate disk. Most of the day the disk is being read at 50k-100k/sec. However at higher questions/sec the disk reads jump up and stay there for on the order of an hour. Running redhay 9 with 2X2.4 xeon with 4GB ram and the active portion of memory (according to top) is ~ 3GB. Filesystem has 3.3 GB and mysql ~500M. According to the ration or key_reads/key_requests the key_buffer is large enough and key_writes/key_write_requests ~ 1. I must admit I don't know if it is the MYI or MYD files that are being read during this period(tried lsof, any other tools to see disk activity by file?). During this period of high read I see one specific query being heavily slowed. Whe shall call this query, A. Query A joins 3 tables but the explain reveals that all use an index with type (const,ref,ref). There is an order by in query A. I thought about extending the size of the sort_buffer or read_rnd_buffer but aren't those just pulling in record data which should be cached by the file system already? I have included the explain of query A below. Any advice would be most appreciated. Regards, Trevor +---+---+--+--+-+---+--+-+ | table | type | possible_keys| key | key_len | ref | rows | Extra | +---+---+--+--+-+---+--+-+ | t | const | PRIMARY,tribe_id_idx | PRIMARY | 40 | const |1 | Using temporary; Using filesort | | s | ref | PRIMARY,statistics_parent_id_idx | statistics_parent_id_idx | 40 | const |1 | Using where | | c | ref | click_statistics_id_idx,click_visitor_id_idx | click_statistics_id_idx | 40 | s.ID | 81 | Using where | +---+---+--+--+-+---+--+-+ 3 rows in set (0.00 sec) Trevor -- MySQL General Mailing List For list archives: http://lists.mysql.com/mysql To unsubscribe:http://lists.mysql.com/[EMAIL PROTECTED]