On Tue, 4 Dec 2001 at 8:54am, Stefan G. Weichinger wrote > JBL> That sounds like a duplex mismatch. Are the 10M boxen on a switch smart > JBL> enough to deal with this? > > Yes thereīs a switch inbetween. > Most of those slow boxes are connected via a hub (coax <-> cat5). > I monitored switch activity, the nic of the amanda machine negotiates > Full Duplex with 100M ... so I did not explicitly lock the nic to > 100M.
I'm wondering if you're seeing lots of collisions when the 10M clients start dumping. > JBL> What is your runspercycle? Tapecycle *must* be >= (preferebably >) > JBL> runspercycle. Otherwise, you may be overwriting the *only* full dump of a > JBL> filesystem you have on tape. If that happens, you're SOL at restore time. > > I donīt have runspercycle set. > Wrong ? Gonna read about it once more. > > dumpcycle 14 > tapecycle 11 > runtapes 1 I believe (taking into account your next mail) that it's dumpcycle that needs the units. For example, I want a full dump of every filesystem once a week, running only on weeknights. So I have dumpcycle 1 week runspercycle 5 runtapes 1 I also want a 3 month history of dailies, so I have tapecycle 60 I'm guessing that what you want is: dumpcycle 2 weeks runspercycle 14 runtapes 1 As I mentioned, then, you need *at least* 14 tapes. 15 or 16 would be *much* better. > btw, whatīs SOL ? The polite version is 'Sorry, Out of Luck'. -- Joshua Baker-LePain Department of Biomedical Engineering Duke University