[JIRA] Created: (LBCLASSIC-264) MDC of parent thread can be wiped from child thread
MDC of parent thread can be wiped from child thread --------------------------------------------------- Key: LBCLASSIC-264 URL: http://jira.qos.ch/browse/LBCLASSIC-264 Project: logback-classic Issue Type: Bug Affects Versions: 0.9.28 Reporter: Iwein Fuld Assignee: Logback dev list In our web framework we're using the MDC, but because we use multiple rendering threads we need to copy context from one thread to another. With log4j this worked fine, with logback it doesn't. It seems that there is an issue at least with setContextMap, which can influence the MDC in other (non child) threads. I've reproduced one such issue in a small project: https://github.com/iwein/logback-testdrive Look at this testcase in particular: {code} @Test public void shouldCopyContextToOtherThread() throws Exception { final AtomicReference<Map> contextMap = new AtomicReference<Map>(); MDC.put("first", "Dorothy"); MDC.put("last", "Parker"); contextMap.set(MDC.getCopyOfContextMap()); //some context is set in a different thread Executors.newSingleThreadExecutor().execute(new Runnable() { public void run() { //not needed, but shouldn't break anything in the parent right? MDC.setContextMap(contextMap.get()); logger.info("Check enclosed."); } }); //If I don't do this I have lost my context. That seems wrong.. //MDC.setContextMap(contextMap.get()); logger.debug("The most beautiful two words in English."); } {code} I'll try to reproduce the issue with some sibling threads for the heck of it, but I think the above already illustrates the problem. -- This message is automatically generated by JIRA. - If you think it was sent incorrectly contact one of the administrators: http://jira.qos.ch/secure/Administrators.jspa - For more information on JIRA, see: http://www.atlassian.com/software/jira
[ http://jira.qos.ch/browse/LBCLASSIC-264?page=com.atlassian.jira.plugin.syste... ] Ceki Gulcu reassigned LBCLASSIC-264: ------------------------------------ Assignee: Ceki Gulcu (was: Logback dev list)
MDC of parent thread can be wiped from child thread ---------------------------------------------------
Key: LBCLASSIC-264 URL: http://jira.qos.ch/browse/LBCLASSIC-264 Project: logback-classic Issue Type: Bug Affects Versions: 0.9.28 Reporter: Iwein Fuld Assignee: Ceki Gulcu
In our web framework we're using the MDC, but because we use multiple rendering threads we need to copy context from one thread to another. With log4j this worked fine, with logback it doesn't. It seems that there is an issue at least with setContextMap, which can influence the MDC in other (non child) threads. I've reproduced one such issue in a small project: https://github.com/iwein/logback-testdrive Look at this testcase in particular: {code} @Test public void shouldCopyContextToOtherThread() throws Exception { final AtomicReference<Map> contextMap = new AtomicReference<Map>(); MDC.put("first", "Dorothy"); MDC.put("last", "Parker"); contextMap.set(MDC.getCopyOfContextMap()); //some context is set in a different thread Executors.newSingleThreadExecutor().execute(new Runnable() { public void run() { //not needed, but shouldn't break anything in the parent right? MDC.setContextMap(contextMap.get()); logger.info("Check enclosed."); } }); //If I don't do this I have lost my context. That seems wrong.. //MDC.setContextMap(contextMap.get()); logger.debug("The most beautiful two words in English."); } {code} I'll try to reproduce the issue with some sibling threads for the heck of it, but I think the above already illustrates the problem.
-- This message is automatically generated by JIRA. - If you think it was sent incorrectly contact one of the administrators: http://jira.qos.ch/secure/Administrators.jspa - For more information on JIRA, see: http://www.atlassian.com/software/jira
[ http://jira.qos.ch/browse/LBCLASSIC-264?page=com.atlassian.jira.plugin.syste... ] Iwein Fuld commented on LBCLASSIC-264: -------------------------------------- Added a sibling test here: https://github.com/iwein/logback-testdrive/commit/5d7a24a44ad4f3be99e794f192... It seems that LogbackMDCAdapter doesn't use CopyOnInheritThreadLocal while the field name suggest that it was intended to do so. If I feel inspired this weekend I might crank out a patch.
MDC of parent thread can be wiped from child thread ---------------------------------------------------
Key: LBCLASSIC-264 URL: http://jira.qos.ch/browse/LBCLASSIC-264 Project: logback-classic Issue Type: Bug Affects Versions: 0.9.28 Reporter: Iwein Fuld Assignee: Ceki Gulcu
In our web framework we're using the MDC, but because we use multiple rendering threads we need to copy context from one thread to another. With log4j this worked fine, with logback it doesn't. It seems that there is an issue at least with setContextMap, which can influence the MDC in other (non child) threads. I've reproduced one such issue in a small project: https://github.com/iwein/logback-testdrive Look at this testcase in particular: {code} @Test public void shouldCopyContextToOtherThread() throws Exception { final AtomicReference<Map> contextMap = new AtomicReference<Map>(); MDC.put("first", "Dorothy"); MDC.put("last", "Parker"); contextMap.set(MDC.getCopyOfContextMap()); //some context is set in a different thread Executors.newSingleThreadExecutor().execute(new Runnable() { public void run() { //not needed, but shouldn't break anything in the parent right? MDC.setContextMap(contextMap.get()); logger.info("Check enclosed."); } }); //If I don't do this I have lost my context. That seems wrong.. //MDC.setContextMap(contextMap.get()); logger.debug("The most beautiful two words in English."); } {code} I'll try to reproduce the issue with some sibling threads for the heck of it, but I think the above already illustrates the problem.
-- This message is automatically generated by JIRA. - If you think it was sent incorrectly contact one of the administrators: http://jira.qos.ch/secure/Administrators.jspa - For more information on JIRA, see: http://www.atlassian.com/software/jira
[ http://jira.qos.ch/browse/LBCLASSIC-264?page=com.atlassian.jira.plugin.syste... ] Iwein Fuld commented on LBCLASSIC-264: -------------------------------------- This issue is fixed in 0.9.29, as collateral for the fix for LBCLASSIC-253
MDC of parent thread can be wiped from child thread ---------------------------------------------------
Key: LBCLASSIC-264 URL: http://jira.qos.ch/browse/LBCLASSIC-264 Project: logback-classic Issue Type: Bug Affects Versions: 0.9.28 Reporter: Iwein Fuld Assignee: Ceki Gulcu
In our web framework we're using the MDC, but because we use multiple rendering threads we need to copy context from one thread to another. With log4j this worked fine, with logback it doesn't. It seems that there is an issue at least with setContextMap, which can influence the MDC in other (non child) threads. I've reproduced one such issue in a small project: https://github.com/iwein/logback-testdrive Look at this testcase in particular: {code} @Test public void shouldCopyContextToOtherThread() throws Exception { final AtomicReference<Map> contextMap = new AtomicReference<Map>(); MDC.put("first", "Dorothy"); MDC.put("last", "Parker"); contextMap.set(MDC.getCopyOfContextMap()); //some context is set in a different thread Executors.newSingleThreadExecutor().execute(new Runnable() { public void run() { //not needed, but shouldn't break anything in the parent right? MDC.setContextMap(contextMap.get()); logger.info("Check enclosed."); } }); //If I don't do this I have lost my context. That seems wrong.. //MDC.setContextMap(contextMap.get()); logger.debug("The most beautiful two words in English."); } {code} I'll try to reproduce the issue with some sibling threads for the heck of it, but I think the above already illustrates the problem.
-- This message is automatically generated by JIRA. - If you think it was sent incorrectly contact one of the administrators: http://jira.qos.ch/secure/Administrators.jspa - For more information on JIRA, see: http://www.atlassian.com/software/jira
participants (2)
-
Ceki Gulcu (JIRA) -
Iwein Fuld (JIRA)