1:53 p.m.
Hello Dev List,
I encountered an issue today with my KnowledgeAgent removing resources
from its RuleBase shortly after creating it. I have the
ResourceChangeScanner running in my application.
I tracked the issue back to the scan() method in
ResourceChangeScannerImpl. It appears that the method is trying to
identify resources that are no longer available and remove them from both
the RuleBase and future scans. To do this it is checking lastModified on
the resource and on a result of 0 removing the resource. The resources
that I configured in my change-set definitely still exist, but due to URL
handler implementation provided by my classloader, getLastModified always
returns 0. (The resource I'm retrieving is coming from a jar that is in
my application's classpath and the URL handler implementation is
oracle.classloader.SharedCodeSourceURL)
Do you think it would be possible for the scan to identify unavailable
resources some other way than with the lastModified? and then if
lastModified is 0 maybe always or never update the resource? I'm not sure
what the best approach to that would be, but removing resources when their
lastModified is 0 seems incorrect to me.
Thanks,
Steve Ronderos
5:29 p.m.
Hi Steve - I have seen that issue before reported by others. But you
just explained the cause ! - I don't often use the classpath scanner
myself, so never thought of that.
Yes the lastModified being zero is a problem...
OK, well I think perhaps the solution would be if lastModified is not
a valid value, to NOT use that to decide availablility.
I think using a scanner on a classpath resource is a bit unusual to
start with, but it is possible in various containers to dynamically
change the classpath, so it kind of technically makes sense.
If I had things in the classpath, I would assume they are constant -
but I assume you are actually expecting the scan to pick up changes in
a future ?
(in the meantime, if you can avoid the classpath one that problem
should go away).
On Thu, Oct 8, 2009 at 5:53 AM, Steve Ronderos <steve.ronderos(a)ni.com> wrote:
Hello Dev List,
I encountered an issue today with my KnowledgeAgent removing resources from
its RuleBase shortly after creating it. I have the ResourceChangeScanner
running in my application.
I tracked the issue back to the scan() method in ResourceChangeScannerImpl.
It appears that the method is trying to identify resources that are no
longer available and remove them from both the RuleBase and future scans.
To do this it is checking lastModified on the resource and on a result of 0
removing the resource. The resources that I configured in my change-set
definitely still exist, but due to URL handler implementation provided by my
classloader, getLastModified always returns 0. (The resource I'm retrieving
is coming from a jar that is in my application's classpath and the URL
handler implementation is oracle.classloader.SharedCodeSourceURL)
Do you think it would be possible for the scan to identify unavailable
resources some other way than with the lastModified? and then if
lastModified is 0 maybe always or never update the resource? I'm not sure
what the best approach to that would be, but removing resources when their
lastModified is 0 seems incorrect to me.
Thanks,
Steve Ronderos
_______________________________________________
rules-dev mailing list
rules-dev(a)lists.jboss.org
https://lists.jboss.org/mailman/listinfo/rules-dev
--
Michael D Neale
home: www.michaelneale.net
blog: michaelneale.blogspot.com
8:46 a.m.
Michael,
In this exact case I am not expecting the scanner to ever find an updated
resource; however, the changeset is configured to read one resource from a
URL and another from the classpath. The resource from the URL is expected
to change, but the one from the classpath is not. We have a use case
where we may need classpath resources to be scanned, but I'm not sure it
will ever come up.
Unfortunately I don't think that we can reference that resource any way
other than the classpath at this moment, but we do have another
workaround. Previous to Drools 5 we had a custom way of refreshing our
resources which I think we will go back to for now.
Should I log a feature request for this change?
Steve Ronderos
rules-dev-bounces(a)lists.jboss.org wrote on 10/07/2009 05:29:56 PM:
[image removed]
Re: [rules-dev] Issue with ResourceChangeScanner
Michael Neale
to:
Rules Dev List
10/07/2009 05:31 PM
Sent by:
rules-dev-bounces(a)lists.jboss.org
Please respond to Rules Dev List
Hi Steve - I have seen that issue before reported by others. But you
just explained the cause ! - I don't often use the classpath scanner
myself, so never thought of that.
Yes the lastModified being zero is a problem...
OK, well I think perhaps the solution would be if lastModified is not
a valid value, to NOT use that to decide availablility.
I think using a scanner on a classpath resource is a bit unusual to
start with, but it is possible in various containers to dynamically
change the classpath, so it kind of technically makes sense.
If I had things in the classpath, I would assume they are constant -
but I assume you are actually expecting the scan to pick up changes in
a future ?
(in the meantime, if you can avoid the classpath one that problem
should go away).
On Thu, Oct 8, 2009 at 5:53 AM, Steve Ronderos <steve.ronderos(a)ni.com>
wrote:
>
> Hello Dev List,
>
> I encountered an issue today with my KnowledgeAgent removing resources
from
> its RuleBase shortly after creating it. I have the
ResourceChangeScanner
> running in my application.
>
> I tracked the issue back to the scan() method in
ResourceChangeScannerImpl.
> It appears that the method is trying to identify resources that
are
no
> longer available and remove them from both the RuleBase and
future
scans.
> To do this it is checking lastModified on the resource and on a
result of 0
> removing the resource. The resources that I configured in my
change-set
> definitely still exist, but due to URL handler implementation
provided
by my
> classloader, getLastModified always returns 0. (The resource
I'm
retrieving
> is coming from a jar that is in my application's classpath
and the URL
> handler implementation is oracle.classloader.SharedCodeSourceURL)
>
> Do you think it would be possible for the scan to identify unavailable
> resources some other way than with the lastModified? and then if
> lastModified is 0 maybe always or never update the resource? I'm not
sure
> what the best approach to that would be, but removing resources
when
their
> lastModified is 0 seems incorrect to me.
>
> Thanks,
>
> Steve Ronderos
> _______________________________________________
> rules-dev mailing list
> rules-dev(a)lists.jboss.org
> https://lists.jboss.org/mailman/listinfo/rules-dev
>
>
--
Michael D Neale
home: www.michaelneale.net
blog: michaelneale.blogspot.com
_______________________________________________
rules-dev mailing list
rules-dev(a)lists.jboss.org
https://lists.jboss.org/mailman/listinfo/rules-dev
5:13 p.m.
Hi Steve - ok - I am leaning towards insisting that classpath
resources are not scanned for - just loaded once. There are cases
where you can update a classpath value on the fly, but really, I think
using file/database/http would be better for that sort of dynamic
changes...
Please do log it as a bug, let us know and I will make it critical, as
I think its kinda broken and not workable at the moment.
On Fri, Oct 9, 2009 at 12:46 AM, Steve Ronderos <steve.ronderos(a)ni.com> wrote:
Michael,
In this exact case I am not expecting the scanner to ever find an updated
resource; however, the changeset is configured to read one resource from a
URL and another from the classpath. The resource from the URL is expected
to change, but the one from the classpath is not. We have a use case where
we may need classpath resources to be scanned, but I'm not sure it will ever
come up.
Unfortunately I don't think that we can reference that resource any way
other than the classpath at this moment, but we do have another workaround.
Previous to Drools 5 we had a custom way of refreshing our resources which
I think we will go back to for now.
Should I log a feature request for this change?
Steve Ronderos
rules-dev-bounces(a)lists.jboss.org wrote on 10/07/2009 05:29:56 PM:
> [image removed]
>
> Re: [rules-dev] Issue with ResourceChangeScanner
>
> Michael Neale
>
> to:
>
> Rules Dev List
>
> 10/07/2009 05:31 PM
>
> Sent by:
>
> rules-dev-bounces(a)lists.jboss.org
>
> Please respond to Rules Dev List
>
> Hi Steve - I have seen that issue before reported by others. But you
> just explained the cause ! - I don't often use the classpath scanner
> myself, so never thought of that.
>
> Yes the lastModified being zero is a problem...
>
> OK, well I think perhaps the solution would be if lastModified is not
> a valid value, to NOT use that to decide availablility.
>
> I think using a scanner on a classpath resource is a bit unusual to
> start with, but it is possible in various containers to dynamically
> change the classpath, so it kind of technically makes sense.
>
> If I had things in the classpath, I would assume they are constant -
> but I assume you are actually expecting the scan to pick up changes in
> a future ?
>
> (in the meantime, if you can avoid the classpath one that problem
> should go away).
>
> On Thu, Oct 8, 2009 at 5:53 AM, Steve Ronderos <steve.ronderos(a)ni.com>
> wrote:
> >
> > Hello Dev List,
> >
> > I encountered an issue today with my KnowledgeAgent removing resources
> > from
> > its RuleBase shortly after creating it. I have the
> > ResourceChangeScanner
> > running in my application.
> >
> > I tracked the issue back to the scan() method in
> > ResourceChangeScannerImpl.
> > It appears that the method is trying to identify resources that are no
> > longer available and remove them from both the RuleBase and future
> > scans.
> > To do this it is checking lastModified on the resource and on a result
> > of 0
> > removing the resource. The resources that I configured in my change-set
> > definitely still exist, but due to URL handler implementation provided
> > by my
> > classloader, getLastModified always returns 0. (The resource I'm
> > retrieving
> > is coming from a jar that is in my application's classpath and the URL
> > handler implementation is oracle.classloader.SharedCodeSourceURL)
> >
> > Do you think it would be possible for the scan to identify unavailable
> > resources some other way than with the lastModified? and then if
> > lastModified is 0 maybe always or never update the resource? I'm not
> > sure
> > what the best approach to that would be, but removing resources when
> > their
> > lastModified is 0 seems incorrect to me.
> >
> > Thanks,
> >
> > Steve Ronderos
> > _______________________________________________
> > rules-dev mailing list
> > rules-dev(a)lists.jboss.org
> > https://lists.jboss.org/mailman/listinfo/rules-dev
> >
> >
>
>
>
> --
> Michael D Neale
> home: www.michaelneale.net
> blog: michaelneale.blogspot.com
>
> _______________________________________________
> rules-dev mailing list
> rules-dev(a)lists.jboss.org
> https://lists.jboss.org/mailman/listinfo/rules-dev
_______________________________________________
rules-dev mailing list
rules-dev(a)lists.jboss.org
https://lists.jboss.org/mailman/listinfo/rules-dev
--
Michael D Neale
home: www.michaelneale.net
blog: michaelneale.blogspot.com
11:35 a.m.
Michael,
I've logged the issue (https://jira.jboss.org/jira/browse/JBRULES-2293). I
agree that classpath resources should not be scanned. That would solve my
immediate problem.
One last note, I still don't think that checking lastModified != 0 to
determine existence is ideal, there could be other scenarios where the
lastModified method is not implemented and returns 0. Is there a better
way to check for resource existence other than lastModified?
Thanks,
Steve Ronderos
rules-dev-bounces(a)lists.jboss.org wrote on 10/08/2009 05:13:17 PM:
[image removed]
Re: [rules-dev] Issue with ResourceChangeScanner
Michael Neale
to:
Rules Dev List
10/08/2009 05:14 PM
Sent by:
rules-dev-bounces(a)lists.jboss.org
Please respond to Rules Dev List
Hi Steve - ok - I am leaning towards insisting that classpath
resources are not scanned for - just loaded once. There are cases
where you can update a classpath value on the fly, but really, I think
using file/database/http would be better for that sort of dynamic
changes...
Please do log it as a bug, let us know and I will make it critical, as
I think its kinda broken and not workable at the moment.
On Fri, Oct 9, 2009 at 12:46 AM, Steve Ronderos <steve.ronderos(a)ni.com>
wrote:
>
> Michael,
>
> In this exact case I am not expecting the scanner to ever find an
updated
> resource; however, the changeset is configured to read one
resource
from a
> URL and another from the classpath. The resource from the URL
is
expected
> to change, but the one from the classpath is not. We have a use
case
where
> we may need classpath resources to be scanned, but I'm not
sure itwill
ever
> come up.
>
> Unfortunately I don't think that we can reference that resource any
way
> other than the classpath at this moment, but we do have another
workaround.
> Previous to Drools 5 we had a custom way of refreshing our
resources
which
> I think we will go back to for now.
>
> Should I log a feature request for this change?
>
> Steve Ronderos
>
> rules-dev-bounces(a)lists.jboss.org wrote on 10/07/2009 05:29:56 PM:
>
>> [image removed]
>>
>> Re: [rules-dev] Issue with ResourceChangeScanner
>>
>> Michael Neale
>>
>> to:
>>
>> Rules Dev List
>>
>> 10/07/2009 05:31 PM
>>
>> Sent by:
>>
>> rules-dev-bounces(a)lists.jboss.org
>>
>> Please respond to Rules Dev List
>>
>> Hi Steve - I have seen that issue before reported by others. But you
>> just explained the cause ! - I don't often use the classpath scanner
>> myself, so never thought of that.
>>
>> Yes the lastModified being zero is a problem...
>>
>> OK, well I think perhaps the solution would be if lastModified is not
>> a valid value, to NOT use that to decide availablility.
>>
>> I think using a scanner on a classpath resource is a bit unusual to
>> start with, but it is possible in various containers to dynamically
>> change the classpath, so it kind of technically makes sense.
>>
>> If I had things in the classpath, I would assume they are constant -
>> but I assume you are actually expecting the scan to pick up changes
in
>> a future ?
>>
>> (in the meantime, if you can avoid the classpath one that problem
>> should go away).
>>
>> On Thu, Oct 8, 2009 at 5:53 AM, Steve Ronderos
<steve.ronderos(a)ni.com>
>> wrote:
>> >
>> > Hello Dev List,
>> >
>> > I encountered an issue today with my KnowledgeAgent removing
resources
>> > from
>> > its RuleBase shortly after creating it. I have the
>> > ResourceChangeScanner
>> > running in my application.
>> >
>> > I tracked the issue back to the scan() method in
>> > ResourceChangeScannerImpl.
>> > It appears that the method is trying to identify resources that
are
no
>> > longer available and remove them from both the RuleBase
and future
>> > scans.
>> > To do this it is checking lastModified on the resource and on a
result
>> > of 0
>> > removing the resource. The resources that I configured in my
change-set
>> > definitely still exist, but due to URL handler
implementation
provided
>> > by my
>> > classloader, getLastModified always returns 0. (The resource I'm
>> > retrieving
>> > is coming from a jar that is in my application's classpath and the
URL
>> > handler implementation is
oracle.classloader.SharedCodeSourceURL)
>> >
>> > Do you think it would be possible for the scan to identify
unavailable
>> > resources some other way than with the lastModified?
and then if
>> > lastModified is 0 maybe always or never update the resource? I'm
not
>> > sure
>> > what the best approach to that would be, but removing resources
when
>> > their
>> > lastModified is 0 seems incorrect to me.
>> >
>> > Thanks,
>> >
>> > Steve Ronderos
>> > _______________________________________________
>> > rules-dev mailing list
>> > rules-dev(a)lists.jboss.org
>> > https://lists.jboss.org/mailman/listinfo/rules-dev
>> >
>> >
>>
>>
>>
>> --
>> Michael D Neale
>> home: www.michaelneale.net
>> blog: michaelneale.blogspot.com
>>
>> _______________________________________________
>> rules-dev mailing list
>> rules-dev(a)lists.jboss.org
>> https://lists.jboss.org/mailman/listinfo/rules-dev
>
> _______________________________________________
> rules-dev mailing list
> rules-dev(a)lists.jboss.org
> https://lists.jboss.org/mailman/listinfo/rules-dev
>
>
--
Michael D Neale
home: www.michaelneale.net
blog: michaelneale.blogspot.com
_______________________________________________
rules-dev mailing list
rules-dev(a)lists.jboss.org
https://lists.jboss.org/mailman/listinfo/rules-dev
5:03 p.m.
Yeah I think if lastModified is not implemented, it can't really be
taken into account.
I really think most of this is due to scanning the classpath - in
other cases it is probably reasonable to expect it to be correct - the
other alternative is to read the whole thing into a buffer, and then
compare hashes (SHA-1 or MD5 would be good enough) of what it has
versus what it just downloaded to see if there is a difference. That
isn't really good for polling, as it is a reasonable amount of
computation to do (at least that is my gut feel) and needless
downloading.
Hope that makes sense !
Michael.
On Sat, Oct 10, 2009 at 3:35 AM, Steve Ronderos <steve.ronderos(a)ni.com> wrote:
Michael,
I've logged the issue (https://jira.jboss.org/jira/browse/JBRULES-2293). I
agree that classpath resources should not be scanned. That would solve my
immediate problem.
One last note, I still don't think that checking lastModified != 0 to
determine existence is ideal, there could be other scenarios where the
lastModified method is not implemented and returns 0. Is there a better way
to check for resource existence other than lastModified?
Thanks,
Steve Ronderos
rules-dev-bounces(a)lists.jboss.org wrote on 10/08/2009 05:13:17 PM:
> [image removed]
>
> Re: [rules-dev] Issue with ResourceChangeScanner
>
> Michael Neale
>
> to:
>
> Rules Dev List
>
> 10/08/2009 05:14 PM
>
> Sent by:
>
> rules-dev-bounces(a)lists.jboss.org
>
> Please respond to Rules Dev List
>
> Hi Steve - ok - I am leaning towards insisting that classpath
> resources are not scanned for - just loaded once. There are cases
> where you can update a classpath value on the fly, but really, I think
> using file/database/http would be better for that sort of dynamic
> changes...
>
> Please do log it as a bug, let us know and I will make it critical, as
> I think its kinda broken and not workable at the moment.
>
> On Fri, Oct 9, 2009 at 12:46 AM, Steve Ronderos <steve.ronderos(a)ni.com>
> wrote:
> >
> > Michael,
> >
> > In this exact case I am not expecting the scanner to ever find an
> > updated
> > resource; however, the changeset is configured to read one resource from
> > a
> > URL and another from the classpath. The resource from the URL is
> > expected
> > to change, but the one from the classpath is not. We have a use case
> > where
> > we may need classpath resources to be scanned, but I'm not sure itwill
> > ever
> > come up.
> >
> > Unfortunately I don't think that we can reference that resource any way
> > other than the classpath at this moment, but we do have another
> > workaround.
> > Previous to Drools 5 we had a custom way of refreshing our resources
> > which
> > I think we will go back to for now.
> >
> > Should I log a feature request for this change?
> >
> > Steve Ronderos
> >
> > rules-dev-bounces(a)lists.jboss.org wrote on 10/07/2009 05:29:56 PM:
> >
> >> [image removed]
> >>
> >> Re: [rules-dev] Issue with ResourceChangeScanner
> >>
> >> Michael Neale
> >>
> >> to:
> >>
> >> Rules Dev List
> >>
> >> 10/07/2009 05:31 PM
> >>
> >> Sent by:
> >>
> >> rules-dev-bounces(a)lists.jboss.org
> >>
> >> Please respond to Rules Dev List
> >>
> >> Hi Steve - I have seen that issue before reported by others. But you
> >> just explained the cause ! - I don't often use the classpath scanner
> >> myself, so never thought of that.
> >>
> >> Yes the lastModified being zero is a problem...
> >>
> >> OK, well I think perhaps the solution would be if lastModified is not
> >> a valid value, to NOT use that to decide availablility.
> >>
> >> I think using a scanner on a classpath resource is a bit unusual to
> >> start with, but it is possible in various containers to dynamically
> >> change the classpath, so it kind of technically makes sense.
> >>
> >> If I had things in the classpath, I would assume they are constant -
> >> but I assume you are actually expecting the scan to pick up changes in
> >> a future ?
> >>
> >> (in the meantime, if you can avoid the classpath one that problem
> >> should go away).
> >>
> >> On Thu, Oct 8, 2009 at 5:53 AM, Steve Ronderos
<steve.ronderos(a)ni.com>
> >> wrote:
> >> >
> >> > Hello Dev List,
> >> >
> >> > I encountered an issue today with my KnowledgeAgent removing
> >> > resources
> >> > from
> >> > its RuleBase shortly after creating it. I have the
> >> > ResourceChangeScanner
> >> > running in my application.
> >> >
> >> > I tracked the issue back to the scan() method in
> >> > ResourceChangeScannerImpl.
> >> > It appears that the method is trying to identify resources that are
> >> > no
> >> > longer available and remove them from both the RuleBase and future
> >> > scans.
> >> > To do this it is checking lastModified on the resource and on a
> >> > result
> >> > of 0
> >> > removing the resource. The resources that I configured in my
> >> > change-set
> >> > definitely still exist, but due to URL handler implementation
> >> > provided
> >> > by my
> >> > classloader, getLastModified always returns 0. (The resource I'm
> >> > retrieving
> >> > is coming from a jar that is in my application's classpath and the
> >> > URL
> >> > handler implementation is oracle.classloader.SharedCodeSourceURL)
> >> >
> >> > Do you think it would be possible for the scan to identify
> >> > unavailable
> >> > resources some other way than with the lastModified? and then if
> >> > lastModified is 0 maybe always or never update the resource? I'm
not
> >> > sure
> >> > what the best approach to that would be, but removing resources when
> >> > their
> >> > lastModified is 0 seems incorrect to me.
> >> >
> >> > Thanks,
> >> >
> >> > Steve Ronderos
> >> > _______________________________________________
> >> > rules-dev mailing list
> >> > rules-dev(a)lists.jboss.org
> >> >
https://lists.jboss.org/mailman/listinfo/rules-dev
> >> >
> >> >
> >>
> >>
> >>
> >> --
> >> Michael D Neale
> >> home: www.michaelneale.net
> >> blog: michaelneale.blogspot.com
> >>
> >> _______________________________________________
> >> rules-dev mailing list
> >> rules-dev(a)lists.jboss.org
> >> https://lists.jboss.org/mailman/listinfo/rules-dev
> >
> > _______________________________________________
> > rules-dev mailing list
> > rules-dev(a)lists.jboss.org
> > https://lists.jboss.org/mailman/listinfo/rules-dev
> >
> >
>
>
>
> --
> Michael D Neale
> home: www.michaelneale.net
> blog: michaelneale.blogspot.com
>
> _______________________________________________
> rules-dev mailing list
> rules-dev(a)lists.jboss.org
> https://lists.jboss.org/mailman/listinfo/rules-dev
_______________________________________________
rules-dev mailing list
rules-dev(a)lists.jboss.org
https://lists.jboss.org/mailman/listinfo/rules-dev
--
Michael D Neale
home: www.michaelneale.net
blog: michaelneale.blogspot.com
5524
days inactive
5528
days old
5 comments
2 participants
participants (2)
-
Michael Neale -
Steve Ronderos