Does anyone have experience of this assertion and what might be causing it? It happens fairly regularly with a particular database, but only during a backup. The database passes both express and full validation. We have also tried restoring a backup that pre-dates the first occurrence (then applying subsequent logs). The only reference to it I can find is in the old NTTP newsgroups, where it was suggested that it might be connected to a VMWare bug - but in this case no virtualization is involved. We are doing a full unload / reload just in case it really is a database corruption - but given that restoring the backup didn't fix it, the validation passes and the other case - it seems increasing unlikely that that is the answer. This will take a while to schedule as the database is about 80GB and is used 24x7. |
"Invalid Access to a NULL image" means that we failed to allocate a cache page when auto-growing the cache - the memory page received back was NULL when checking it, and we assert. We have seen this from a few customers and yes, most of those have been associated with virtualized environments. I don't think we've yet had a reproducible in-house of the behaviour to diagnose the memory conditions precisely. Most of our customers were able to remotely work-around the issue by strictly setting the "-ch" parameter on the server-start up switch to an explicit value - does that also work around the behaviour for yourself? Thanks Jeff - that is really useful. I will try that. This is a dedicated server so we could try a fixed cache - would that address the same issue?
(07 Oct '13, 10:40)
Justin Willey
Replies hidden
2
To the best of our understanding, yes, turning off auto-sizing of the cache and fixing the cache size should also work here. Let us know if that works for you.
(07 Oct '13, 10:58)
Jeff Albion
Thanks again - I'll let you know how we get on.
(07 Oct '13, 12:22)
Justin Willey
2
Fixing the cache size seems to have done the trick - no assertions since then. - Many thanks. (for anyone else coming across this - we used -c to set the initial cache size and -ca 0 to prevent re-sizing)
(11 Oct '13, 12:57)
Justin Willey
|
Only two google hits https://www.google.com/#q=%22invalid+access+to+a+null+image%22
...one unavailable right now (probably the one you found), the other on a Cisco forum says "corrupt database". As such, it's pretty rare, so you probably won't find anyone with experience on this forum. Try calling tech support because they have a vast internal problem report database.