On 20-02-2008 12:54:25 +0100, Michael Schmidt wrote:
Hi Fabian,
thanks for your response. I tried to install the lates branch, but get a compile error. The latest nightly build works fine, but does not yet contain the changes, so I will probably wait until tomorrow evening.
Unfortunately I had to do a lot of cross work, so all fixes are in stable only, whereas the feature you want are in the current branch, which doesn't include the fixes (yet).
OK, am I right that all changes will be included in the coming nightly build. If so, I will wait until tomorrow and try again with '-w' flag and your new fixes.
As far as I can see, the fixes from stable have not yet been propagated to the current (HEAD) branch yet. In that respect I am sorry to tell you that you still have to wait for it.
Yes, your script looks fine. You could add some monetdb create bla || die "Failed to create bla" kind of structures to catch problems, but if everything succeeds then you script should work.
Yes, but the problem is not database creation. Rather than that, mserver5 crashes /not before/ loading data into the database. Before connecting to mserver5 (i.e. by loading the sql file), everything looks fine, i.e. mserver5 appears in the process list (and I guess monetdb status indicates that the server is fine).
Your log shows an interesting pattern. It just looks like the database on the even startups sees some left over garbage from the previous run (e.g. an logfile). However, that should not be possible, as destroy is supposed to throw that away as well.
I totally agree. This is why I initially thought it might be a port allocation problem. I tried waiting for ports to be freed again, and also tried timeouts of about 10mins between two consecutive runs, but still the same behavior. There must be some kind of garbage that causes mserver5 to crash as soon as a connection/any other action is triggered.
I'm affraid I can't answer this right away, we will have to reproduce this to see what's going on exactly.
It seems that this error is independent from the data files being loaded into the DB, but if you need some of the data files I am testing with let me know!
Niels thinks that the crashes are due to the sql_logs not being thrown away. Could you check if the var/MonetDB5/sql_logs/DBNAME is removed by monetdb destroy -f as well as var/MonetDB5/dbfarm/DBNAME?