Hi. Just ran into this bug as well (Java 1.5, logback 1.0.0, Windows 7). It appears that the code in SMTPAppenderBase if (eventEvaluator.evaluate(eventObject)) { // perform actual sending asynchronously SenderRunnable senderRunnable = new SenderRunnable(new CyclicBuffer<E>(cb), eventObject); context.getExecutorService().execute(senderRunnable); } is properly executed, but as noted the SenderRunable.run() method is not invoked (which I have determined by having a break point). The ThreadPoolExecutor.execute() method (under Oracle Java 1.5) executes: public void execute(Runnable command) { if (command == null) throw new NullPointerException(); for (;;) { if (runState != RUNNING) { reject(command); return; } if (poolSize < corePoolSize && addIfUnderCorePoolSize(command)) return; if (workQueue.offer(command)) return; and returns (and the SenderRunnable was added to workQueue). There is, however, a potential problem of mails being lost. In the code causing me to investigate this, the "log.error("....", e)" is immediately followed by a System.exit(). In other words - I will most likely loose most of the mails generated by this code as the async process is slower than the System.exit(). Is there any way of guaranteeing that all mail has been transported before stopping the program? If not, can I have a flag simply saying "do not run async, run in process" which invokes senderRunnable.run() directly instead of delegating to an executor? (also a way of telling the operator that mail has been sent would be nice. It can be as simple as a print statement, similar to @Override protected void sendBuffer( ch.qos.logback.core.helpers.CyclicBuffer<ILoggingEvent> cb, ILoggingEvent lastEventObject) { System.out.println("[" + new java.util.Date() + " " + this.getClass().getSimpleName() + " sending e-mail notification with " + cb.length() + " events]"); super.sendBuffer(cb, lastEventObject); }; ) Thanks /Thorbjørn -----Original Message----- From: logback-user-bounces@qos.ch [mailto:logback-user-bounces@qos.ch] On Behalf Of ceki Sent: 29. januar 2012 18:49 To: logback users list Subject: Re: [logback-user] SMTPAppender mail not sent (logback 1.0.0, JDK 1.5) Hi, I intend to look into this tomorrow morning. -- Ceki http://twitter.com/#!/ceki On 29.01.2012 17:36, samyem wrote:

I am facing the exact same issue. The upgrade stopped sending emails for me as well. For some reason, the Runnable class SMTPAppender's "run" method never gets executed in this new version. Anyone else having this issue?

Thanks,

Louis-Félix wrote:

...

Hi,

I upgraded to logback 1.0.0 in a web application (tomcat 5, JDK 1.5), and the ERROR email are not sent anymore from the SMTPAppender. When I rollback my logback-core and logback-classic JARs to version 0.9.30, it's working again.

I have a simple config:

<appender name="courriel" class="ch.qos.logback.classic.net.SMTPAppender"> <smtpHost>smtp.xxx.xx.xx</smtpHost> <to>xxx@xxx.xx.xx</to> <from>no-reply.ti@xxx.xx.xx</from> <subject>Test</subject> <cyclicBufferTracker class="ch.qos.logback.core.spi.CyclicBufferTrackerImpl"> <bufferSize>25</bufferSize> </cyclicBufferTracker> <layout class="ch.qos.logback.classic.html.HTMLLayout"> <pattern>%date%level%thread%logger%line%message</pattern> </layout> </appender>

I am using logback 1.0.0 with no problem in an other project (with JDK 1.6). Is there any known compatibility problem between logback 1.0.0 and JDK 1.5?

Thanks, Louis-Félix

_______________________________________________ Logback-user mailing list Logback-user@qos.ch http://mailman.qos.ch/mailman/listinfo/logback-user

Regarding the early exit code, I believe your runnables are too shortlived to prove anything here. Try putting a 1-2 second delay in before the print statement to more accurately reflect that sending takes time. I still would like just a "send synchroneous, please" option. Having the email sent is more important to us than an occasional slow log statement. /Thorbjørn -----Original Message----- From: logback-user-bounces@qos.ch [mailto:logback-user-bounces@qos.ch] On Behalf Of ceki Sent: 5. marts 2012 18:55 To: logback users list Subject: Re: [logback-user] SMTPAppender mail not sent (logback 1.0.0, JDK 1.5) Hi Endre, I was not able to make headway regarding this issue. Several people suffer from it as attested by LBCLASSIC-323. Regarding the early exit problem, the following code import ch.qos.logback.core.Context; import ch.qos.logback.core.ContextBase; public class Main { public static void main(String[] args) { Context context = new ContextBase(); for(int i = 0; i < 3; i++) { SenderRunnable senderRunnable = new SenderRunnable(""+i); context.getExecutorService().execute(senderRunnable); } System.out.println("done"); System.exit(0); } static class SenderRunnable implements Runnable { String id; SenderRunnable(String id) { this.id = id; } public void run() { System.out.println("SenderRunnable " +id); } } } yields done SenderRunnable 0 SenderRunnable 1 SenderRunnable 2 I am continuing to investigate. -- Ceki http://twitter.com/#!/ceki On 31.01.2012 13:35, Thorbjørn Ravn Andersen wrote:

Hi.

Just ran into this bug as well (Java 1.5, logback 1.0.0, Windows 7).

It appears that the code in SMTPAppenderBase

if (eventEvaluator.evaluate(eventObject)) { // perform actual sending asynchronously SenderRunnable senderRunnable = new SenderRunnable(new CyclicBuffer<E>(cb), eventObject); context.getExecutorService().execute(senderRunnable); }

is properly executed, but as noted the SenderRunable.run() method is not invoked (which I have determined by having a break point).

The ThreadPoolExecutor.execute() method (under Oracle Java 1.5) executes:

public void execute(Runnable command) { if (command == null) throw new NullPointerException(); for (;;) { if (runState != RUNNING) { reject(command); return; } if (poolSize< corePoolSize&& addIfUnderCorePoolSize(command)) return; if (workQueue.offer(command)) return;

and returns (and the SenderRunnable was added to workQueue).

There is, however, a potential problem of mails being lost. In the code causing me to investigate this, the "log.error("....", e)" is immediately followed by a System.exit(). In other words - I will most likely loose most of the mails generated by this code as the async process is slower than the System.exit().

Is there any way of guaranteeing that all mail has been transported before stopping the program?

If not, can I have a flag simply saying "do not run async, run in process" which invokes senderRunnable.run() directly instead of delegating to an executor?

(also a way of telling the operator that mail has been sent would be nice. It can be as simple as a print statement, similar to

@Override protected void sendBuffer( ch.qos.logback.core.helpers.CyclicBuffer<ILoggingEvent> cb, ILoggingEvent lastEventObject) {

System.out.println("[" + new java.util.Date() + " " + this.getClass().getSimpleName() + " sending e-mail notification with " + cb.length() + " events]");

super.sendBuffer(cb, lastEventObject); }; )

Thanks

/Thorbjørn

-----Original Message----- From: logback-user-bounces@qos.ch [mailto:logback-user-bounces@qos.ch] On Behalf Of ceki Sent: 29. januar 2012 18:49 To: logback users list Subject: Re: [logback-user] SMTPAppender mail not sent (logback 1.0.0, JDK 1.5)

Hi,

I intend to look into this tomorrow morning.

-- Ceki http://twitter.com/#!/ceki

On 29.01.2012 17:36, samyem wrote:

...

I am facing the exact same issue. The upgrade stopped sending emails for me as well. For some reason, the Runnable class SMTPAppender's "run" method never gets executed in this new version. Anyone else having this issue?

Thanks,

Louis-Félix wrote:

...

Hi,

I upgraded to logback 1.0.0 in a web application (tomcat 5, JDK 1.5), and the ERROR email are not sent anymore from the SMTPAppender. When I rollback my logback-core and logback-classic JARs to version 0.9.30, it's working again.

I have a simple config:

<appender name="courriel" class="ch.qos.logback.classic.net.SMTPAppender"> <smtpHost>smtp.xxx.xx.xx</smtpHost> <to>xxx@xxx.xx.xx</to> <from>no-reply.ti@xxx.xx.xx</from> <subject>Test</subject> <cyclicBufferTracker class="ch.qos.logback.core.spi.CyclicBufferTrackerImpl"> <bufferSize>25</bufferSize> </cyclicBufferTracker> <layout class="ch.qos.logback.classic.html.HTMLLayout"> <pattern>%date%level%thread%logger%line%message</pattern> </layout> </appender>

I am using logback 1.0.0 with no problem in an other project (with JDK 1.6). Is there any known compatibility problem between logback 1.0.0 and JDK 1.5?

Thanks, Louis-Félix

_______________________________________________ Logback-user mailing list Logback-user@qos.ch http://mailman.qos.ch/mailman/listinfo/logback-user

On 19.04.2012 00:02, Dave wrote:

Hi everyone. First time poster. I thought I'd share my experiences and findings on the SMTPAppender issues.

I've been struggling to get the STMPAppender working consistently ever since version 0.9.29. I've noticed on this mailing list that it might have been related to JDK 1.5. So I've upgraded our environment to JDK 1.6 (along with upgrading logback to 1.0.1) and every seemed to be working fine.

Then a few days ago, I didn't receive an e-mail from a STMPAppender event that I knew for sure I was supposed to get. I isolated and reproduced the event 10 consecutive times, with about 2 seconds between events. When I checked my email log, I only received 9 messages.

I ran the test again, and got all 10 messages as expected. Tried again once more, and again got 10 messages. I decided to run my test one more time before I put it to rest, but this time I got 9 messages.

After browsing the mailing lists for this issue, I saw Ceki's sample code (posted March 5) testing the ExecutorService. I clipped it out and tried it on my machine (running jdk1.6.0_31). I've found that this too was giving inconsistent results. Out of the three processes it suppose to execute, I get results of 3,3,1,0,0,3,3,1,3 .... I was surprised to see it not execute any processes at times. When running on JDK 1.5, I'm unable to get any results.

As you mentioned, setting ThreadPoolExecutor's corePoolSizeto 0 does not work in JDK 1.5. Thank you for investigating this issue. As for incomplete transmission under JDK 1.6, I am unable to reproduce the problem unless System.exit() is invoked to exit the program. Here is the application I am using. package ch.qos.logback.classic.issue.lbclassic232; import ch.qos.logback.core.Context; import ch.qos.logback.core.ContextBase; public class Barebones { public static void main(String[] args) { Context context = new ContextBase(); for(int i = 0; i < 3; i++) { SenderRunnable senderRunnable = new SenderRunnable(""+i); context.getExecutorService().execute(senderRunnable); } System.out.println("done"); //System.exit(0); } static class SenderRunnable implements Runnable { String id; SenderRunnable(String id) { this.id = id; } public void run() { try { Thread.sleep(2000); } catch (InterruptedException e) { } System.out.println("SenderRunnable " +id); } } } Now that the problem affecting JDK 1.5 has been identified a fix on the way. This should fix http://jira.qos.ch/browse/LBCLASSIC-323

Anyways, that's my 0.02 cents that I can offer on this. If there's any other help I can offer pleas let me know

Dave, thank you for investigating this problem. Much appreciated.

-dave -- Ceki http://twitter.com/#!/ceki