Question Re: Assertion and '-ufd' behaviour in Technology Q&A

Assertion and '-ufd' behaviour

Former Member — Fri, 26 Oct 2018 05:04:47 GMT

[SQL Anywhere 17.0.8.4103]

We have a single engine (dbsrv17.exe) running multiple databases. Recently we had an assertion error, with the following text appearing in the log:

E. 10/25 19:33:13. ERROR Assertion failed: 200505 (17.0.8.4103) Checksum failure on page 10798

We are not using the '-ufd' switch so based on the help file I would expect the default 'abort' value to be in place i.e. 'The affected database is stopped. The status of the database server and other databases remains unchanged.' However, what actually happened is that the engine stopped (the dbsrv17.exe process ended). It was possible to restart the engine without an assertion but I suspect at some point it will happen again.

Some questions on this:

1) Have I misunderstood the nature of this assertion and it is actually related to the engine rather than a single database?

2) Or is the default value for -ufd not working as documented?

3) Is there any way of finding out from the engine what it thinks the current -ufd value is (some undocumented server or database property)?

4) Assuming that the assertion is related to a database rather than the engine, is there any easy way of finding out which database is the cause? ('easy' being some solution other than trying to divide the databases between multiple engines and see which one fails or running dbvalid across all databases)

Re: Assertion and '-ufd' behaviour

Breck_Carter — Sun, 28 Oct 2018 11:43:49 GMT

Sounds like a bug but... since it is difficult ( ok, impossible for me 🙂 to test -ufd behavior, a call to tech support may be your best approach

Re: Assertion and '-ufd' behaviour

VolkerBarth — Mon, 29 Oct 2018 04:16:58 GMT

There are a few hits related to that assertion in SQL Anywhere forums but they often show a database name as part of the message, such as

ERROR Assertion failed: 200505(12.0.1.xxxx)[DBNAME]

If there is not database name as part of the message, I could imagine that a server-based error has appeared (I guess the server stores internal information in pages, too.)

But note, that is a veeery wild guess.

Re: Assertion and '-ufd' behaviour

Former Member — Mon, 29 Oct 2018 05:15:52 GMT

I've subsequently found out that when the assertion occurs the engine crashes rather than shuts down gracefully so I'm guessing the actual behaviour of -ufd is not relevant here. I'll open a case with SAP to have a look at the crash dumps.