> -----Original Message----- > From: Steve Ellis [mailto:el...@brouhaha.com] > Sent: Thursday, January 20, 2011 10:39 AM > To: bacula-users@lists.sourceforge.net > Subject: Re: [Bacula-users] SD Losing Track of Pool > > On 1/20/2011 7:18 AM, Peter Zenge wrote: > >> > >>> Second, in the Device Status section at the bottom, the pool of LF- > F- > >> 0239 is > >>> listed as "*unknown*"; similarly, under "Jobs waiting to reserve a > >> drive", > >>> each job wants the correct pool, but the current pool is listed as > >> "". > >> > > Admittedly I confused the issue by posting an example with two Pools > involved. Even in that example though, there were jobs using the same > pool as the mounted volume, and they wouldn't run until the 2 current > jobs were done (which presumably allowed the SD to re-mount the same > volume, set the current mounted pool correctly, and then 4 jobs were > able to write to that volume concurrently, as designed. > > > > I saw this issue two other times that day; each time the SD changed > the mounted pool from "LF-Inc" to "*unknown*" and that brought > concurrency to a screeching halt. > > > > Certainly I could bypass this issue by having a dedicated volume and > device for each backup client, but I have over 50 clients right now and > it seems like that should be unnecessary. Is that what other people > who write to disk volumes do? > I've been seeing this issue myself--it only seems to show up for me if > a > volume change happens during a running backup. Once that happens, > parallelism using that device is lost. For me this doesn't happen too > often, as I don't have that many parallel jobs, and most of my backups > are to LTO3, so volume changes don't happen all that often either. > However, it is annoying. > > I thought I had seen something that suggested to me that this issue > might be fixed in 5.0.3, I've recently switched to 5.0.3, but haven't > seen any pro or con results yet. > > On a somewhat related note, it seemed to me that during despooling, all > other spooling jobs stop spooling--this might be intentional, I > suppose, > but I think my disk subsystem would be fast enough to keep up one > despool to LTO3, while other jobs could continue to spool--I could > certainly understand if no other job using the same device was allowed > to start despooling during a despool, but that isn't what I observe. > > If my observations are correct, it would be nice if this was a > configurable choice (with faster tape drives, few disk subsystems would > be able to handle a despool and spooling at the same time)--some of my > jobs stall long enough when this happens to allow some of my desktop > backup clients to go to standby--which means those jobs will fail (my > backup strategy uses Wake-on-LAN to wake them up in the first place). > I > certainly could spread my jobs out more in time, if necessary, to > prevent this, but I like for the backups to happen at night when no one > is likely to be using the systems for anything else. I guess another > option would be to launch a keepalive WoL script when a job starts, and > arrange that the keepalive program be killed when the job completes. > > -se >
Agree about the volume change. In fact I'm running a backup right now that should force a volume change in a couple of hours, and I'm watching the SD status to see if the mounted pool becomes unknown around that time. I have certainly noticed that long-running jobs seem to cause this issue, and it occurred to me that long-running jobs also have a higher chance of spanning volumes. If that's what I see, then I will upgrade to 5.0.3. I can do that pretty quickly, and will report back... ------------------------------------------------------------------------------ Protect Your Site and Customers from Malware Attacks Learn about various malware tactics and how to avoid them. Understand malware threats, the impact they can have on your business, and how you can protect your company and customers by using code signing. http://p.sf.net/sfu/oracle-sfdevnl _______________________________________________ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users