upgrade 1.8 to 2.0: The Need for Speed
No no, not talking about this 
since ZBX-5125 I've been able to update zbx to 2.0
This is done on a VM with a copy of the database. The DB itself has about 15 million records in history and trends, and about 4 million records in events.
The update script has been running for almost 11 days now.
It has been processing this query:
And when I look at the updatescript, another query just as big as this one is coming next...
These queries are running on a single cpu core (on a multicore VM guest), taking up 100% cpu on that core. hardly any disk activity, the load is purely single-cpu-core-bound
Now I have 4 options:
1) sit back, relax, good work takes some time, Rome wasn't built in one day...blablabla
2) Buy myself an intel core i7 3770Kwith a phase change cooling system, overclock the hell out of it, let my boss pay for it and look if it will go any faster than before (so far I think this solution is by far the best somehow
)
3) truncate the large tables and go on with only the base tables, perhaps keep the old db available as an archive for the next couple of months
4) see if things can be sped up softwarewise. For example, it might be possible to do part of the updatescript, or at least, part of this, with tools like sed, awk, grep, cut and tr to an export of the database?
About the 11 days running: I havent seen any progress in the tmpdir directory, newest file in there is 5 days old, no more new entries in that file as well
strace shows nothing going on:
left it at that a couple of minutes.
Ideas welcome.
<edit>
I cancelled the upgrade after almost 15 days. Still no new files / rows in files in mysql tmpdir, still the same query running(?)
Will try with emptied events table.
since ZBX-5125 I've been able to update zbx to 2.0
This is done on a VM with a copy of the database. The DB itself has about 15 million records in history and trends, and about 4 million records in events.
The update script has been running for almost 11 days now.
It has been processing this query:
Code:
SELECT e1.eventid,(SELECT e2.value
FROM events e2
WHERE e2.source=e1.source
AND e2.object=e1.object
AND e2.objectid=e1.objectid
AND (e2.clock<e1.clock OR (e2.clock=e1.clock AND e2.eventid<e1.eventid))
AND e2.value<2
ORDER BY e2.source DESC,
e2.object DESC,
e2.objectid DESC,
e2.clock DESC,
e2.eventid DESC,
e2.value DESC
LIMIT 1) AS prev_value
FROM events e1
WHERE e1.source=0
AND e1.object=0
AND e1.value=0
HAVING prev_value IS NULL OR prev_value = 1
These queries are running on a single cpu core (on a multicore VM guest), taking up 100% cpu on that core. hardly any disk activity, the load is purely single-cpu-core-bound
Now I have 4 options:
1) sit back, relax, good work takes some time, Rome wasn't built in one day...blablabla
2) Buy myself an intel core i7 3770Kwith a phase change cooling system, overclock the hell out of it, let my boss pay for it and look if it will go any faster than before (so far I think this solution is by far the best somehow
)3) truncate the large tables and go on with only the base tables, perhaps keep the old db available as an archive for the next couple of months
4) see if things can be sped up softwarewise. For example, it might be possible to do part of the updatescript, or at least, part of this, with tools like sed, awk, grep, cut and tr to an export of the database?
About the 11 days running: I havent seen any progress in the tmpdir directory, newest file in there is 5 days old, no more new entries in that file as well
strace shows nothing going on:
Code:
root@zbxtst:~# strace -p 1431 Process 1431 attached - interrupt to quit select(14, [12 13], NULL, NULL, NULL
Ideas welcome.
<edit>
I cancelled the upgrade after almost 15 days. Still no new files / rows in files in mysql tmpdir, still the same query running(?)
Will try with emptied events table.
Comment