One other thing about chains of KahaDB files: they can only occur when you've got unacked messages. If you have a chain of files, and you think you have no unacked messages, then you're wrong and you need to figure out where the message actually is. You sent some emails yesterday saying you've got KahaDB journal chains and some saying txlog was your issue, and it wasn't clear if the later messages meant "I was mistaken, there's no chain after all" or "we have both chains of journal files and large txlog files", but if you have chains of journal files, you probably don't have zero unconsumed messages.
Tim On Wed, Sep 30, 2015 at 8:42 AM, Tim Bain <tb...@alumni.duke.edu> wrote: > If anything, you want smaller files not bigger ones, because you're hoping > for a situation where for a given file, all of the messages in the file > have their acks in the same file. If that happens (and it's pure luck of > the draw), you break the chain of having to keep newer files because they > ack messages in older files, and you'll then be able to delete the later > files once all their messages are consumed. The more boundaries between > files there are, the more places where you have a shot at that happening, > so more smaller files increases your odds (but never guarantees anything). > > But you're right that all of the above doesn't sound like a solution to > your problem, because it sounds like it's your txlog that's your problem, > not the KahaDB journal files. > > Unfortunately I know nothing at all about those files (I've never used > transactions with ActiveMQ), but I know others on here have and hopefully > one of them will have some advice to give. > > On Tue, Sep 29, 2015 at 11:59 AM, <barry.barn...@wellsfargo.com> wrote: > >> Why would the txlog (transaction log) be constantly growing with no >> messages on the queues? This is taking up all of our disk space, not the >> kahadb journal files! >> >> Are there any howl log settings we should be using aside from the >> defaults? >> >> Regards, >> >> Barry >> >> >> -----Original Message----- >> From: Christopher Shannon [mailto:christopher.l.shan...@gmail.com] >> Sent: Tuesday, September 29, 2015 11:34 AM >> To: users@activemq.apache.org >> Subject: Re: Nothing on queues, Journal files not getting cleaned up >> >> There is no option to compact KahaDB. The current solution for the >> problem of journal files not being cleaned up is to use multi kahadb. Gary >> has a good post about it here >> http://blog.garytully.com/2011/11/activemq-multiple-kahadb-instances.html >> >> There's a couple open tickets to add compaction to KahaDB. >> >> On Tue, Sep 29, 2015 at 10:10 AM, Clebert Suconic < >> clebert.suco...@gmail.com >> > wrote: >> >> > Isn't there any option to compact the journals on activemq5? >> > >> > In a journal sometimes you have the add-delete-add-add-delete somehow >> > that you can't can't really clean it unless you compact the whole >> > thing. Most of the times I had this issue was because I had a pending >> TX or something. >> > >> > >> > Also: is there any pending XA Transactions? >> > >> > On Tue, Sep 29, 2015 at 9:29 AM, <barry.barn...@wellsfargo.com> wrote: >> > >> > > All Topic queue sizes in JConsole show 0, and dispatch counts show 0. >> > Any >> > > other suggestions? >> > > >> > > Regards, >> > > >> > > Barry >> > > -----Original Message----- >> > > From: tbai...@gmail.com [mailto:tbai...@gmail.com] On Behalf Of Tim >> > > Bain >> > > Sent: Tuesday, September 29, 2015 1:14 AM >> > > To: ActiveMQ Users >> > > Subject: Re: Nothing on queues, Journal files not getting cleaned up >> > > >> > > I trust you've confirmed that there are no messages pending on >> > > topics as well as queues? JMX is the easiest way I know to confirm >> > > that if you haven't already done so. >> > > >> > > Are some but not all journal files being deleted, or none at all? >> > > On Sep 28, 2015 8:44 AM, <barry.barn...@wellsfargo.com> wrote: >> > > >> > > > Hello, >> > > > >> > > > How is it that our journal files continue to grow, yet no messages >> > > > are >> > on >> > > > the queues? >> > > > >> > > > ActiveMQ 5.11.1, Using KahaDB (not levelDB). >> > > > >> > > > How can you tell if a kahaDB scheduler is waiting, and in need of >> > > clearing? >> > > > >> > > > Nothing on the Advisory queues (no network of brokers), nothing on >> > > > the DLQ. Where else can I look? Our txlogs are very large as well. >> > > > Application Teams advise their transaction processing IS >> > > > committing on >> > > gets >> > > > and puts, etc.. >> > > > >> > > > Regards, >> > > > >> > > > Barry >> > > > >> > > > >> > > > >> > > >> > >> > >> > >> > -- >> > Clebert Suconic >> > >> > >
