On 2024-08-16 08:32, Ben Rubinstein via use-livecode wrote:
I have a tool (a LiveCode standalone) running on Windows, which every
night drops and recreates a database on a remote MySQL server, (about
350MB, 50 tables). Running for many years.
About a year ago, we started to see a problem where sometimes the
nightly build would fail, part-way through the process. The routine
involves creating and populating tables, then creating indexes.
Depending when the problem hits, the initial error is either
Connection was killed
or
Lost connection to MySQL server during query
All subsequent calls to revdb_execute get the error
MySQL server has gone away
I tried splitting the build into sections, so that the code opens the
connection to a database builds some of the tables, then closes the
connection, and opens a new connection to add more tables. There was no
evidence that this made the issue occur less frequently; and once it
hit, subsequent attempts to open a connection would get the error
Can't connect to MySQL server on '<address>' (0)
When this was happening maybe a couple of times per month (on average)
it didn't matter too much (the system is designed to be resilient, if
the data wasn't refreshed one day, it would be the next). I thought it
might be network glitches.
Recently IT tightened security on the machine where the tool runs; and
since then we get this problem nine times of out ten. They say the only
change made was to remove the admin privileges of the user account, and
have now reversed that change; however, this problem has remained
since. Another problem that arrived at the same time, reported here as
"a windows weirdness", seems (per Paul Dupuis and Mark Waddingham) to
be related to UNC paths, and possibly to security policies.
Does anyone have a suggestion for how conditions could affect this? Is
there any way to get more detailed information out of the rev database
driver about what's happening?
The dbmysql is just a thin wrapper around the mysqlclient library which
is basically just implementing a protocol over a socket.
The fact that you get 'Can't connect to MySQL on '<address>'' after it
happens sounds very much like there's some sort of blocking going on at
the system level. (A bit like most servers have 'portsentry' or similar
on it which blocks requests which look dodgy) - given the 'tigtening of
security' this is quite possible...
I think there are some low-level network tools on windows you could use
to look at what's happening with sockets/ports (e.g.
https://learn.microsoft.com/en-us/sysinternals/downloads/tcpview).
I was going to suggest tweaking the timeouts/auto-reconnect parameters
on your revOpenDatabase call - but I'm not sure its a timeout due to the
'Can't connect' error happening subsequently.
Warmest Regards
Mark.
--
Mark Waddingham ~ m...@livecode.com ~ http://www.livecode.com/
LiveCode: Build Amazing Things
_______________________________________________
use-livecode mailing list
use-livecode@lists.runrev.com
Please visit this url to subscribe, unsubscribe and manage your subscription
preferences:
http://lists.runrev.com/mailman/listinfo/use-livecode
