Hi, >>>> 5.1.30 is going into OpenSolaris as I type. Later 5.1 releases >>>> will >>>> come as resources allow and criticality necessitates. >>> >>> Thanks Lars! One related question... any chances we have our >>> delta patches from Cool Stack which had a MySQL DTrace provider in >>> this putback? Or for that matter, any DTrace provider patches? >>> >> >> >> I ported a slightly different set of probes back into the 5.1.30 >> drop. There are a few more than those detailed in the documentation >> I wrote here: >> >> http://dev.mysql.com/doc/refman/6.0/en/dba-dtrace-server.html > > Excellent! Thanks much for the info. I know of a number of folks > who will be interested in this. We should work to make sure it's > documented for Web Stack 1.5 > > Looking there, it appears one difference is no direct access to the > THD structure. In the past, I'd found many of the things I wanted > to do needed to follow this around MySQL. Is the intent with the > connectionid and query to be able to follow this around? > Maybe I'm misreading this though.
No, you are not misreading. The connectionid we expose at different levels is the same as the MySQL process ID (as shown within MySQL with SHOW PROCESSLIST), and we also provide the dbname, user, and host at specific levels where appropriate. The most common reason for needing access to the THD structure is either to get a common reference number number (the thread process ID as used in MySQL) and/or detail like the database and user that submitted the query. Exposing the THD is messy, consumes a lot of resource (comparatively) when tracing and has the unfortunate effect of exposing a lot of information that most people either don't need or shouldn't need to have access to. If there's something else you want from THD that isn't there now, or isn't available within the probe you want to use, let me know. I am always to suggestion on the DTrace probes. We are working internally to make them useful not only to end users but also developers of MySQL to help understand different parts of the system. As a side note, I know that in the original Webstack DTrace patch the probes were more detailed for some operations (checking the old patch log flushing, and InnoDB wait buffer and index walking). I'll make sure to add these to the list of future probes. MC -- Martin 'MC' Brown, mc at mcslp.com and mc.brown at sun.com Technical Writer, Database Group, Sun Microsystems Everything MCslp: http://planet.mcslp.com
