[jira] [Commented] (TS-948) do not reload bad remap.config
[ https://issues.apache.org/jira/browse/TS-948?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13216524#comment-13216524 ] Brian Geffon commented on TS-948: - Backported to 3.0.x in commit 2f9ccb0e21cdcf8392ee15f909f2b12d8b35d991 do not reload bad remap.config -- Key: TS-948 URL: https://issues.apache.org/jira/browse/TS-948 Project: Traffic Server Issue Type: Improvement Components: Configuration, Management Affects Versions: 3.1.0, 3.0.1 Reporter: Conan Wang Assignee: Leif Hedstrom Priority: Critical Fix For: 3.1.1 traffic_server will exit if exists bad rules in remap.config, whenever startup or reload ( traffic_line -x ). If remap.config is not edited correctly, the server/cluster will crash when reloading. It's better to pre-check the remap table and not switch to the new table if remap.config is not enough correct. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (TS-948) do not reload bad remap.config
[
https://issues.apache.org/jira/browse/TS-948?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13128090#comment-13128090
]
Conan Wang commented on TS-948:
---
My test:
return TS_ERROR directly in TSRemapInit
{code}
TSReturnCodeâ‹…
TSRemapInit(TSRemapInterface* api_info, char* errbuf, int errbuf_size)
{
return TS_ERROR;
{code}
{code}
[Oct 15 14:51:28.411] Server {0x7fff739b6960} WARNING: Failed to initialize
plugin /Users/conan/box/ts-trunk/libexec/trafficserver/add_header.so (non-zero
retval) ... bailing out
FATAL: UnixUDPNet.cc:295: failed assert `event != NULL`
/Users/conan/box/ts-trunk/bin/traffic_server - STACK TRACE:
0 libtsutil.3.dylib 0x000105af6e3b ink_fatal_va + 283
1 libtsutil.3.dylib 0x000105af7144 ink_fatal + 356
2 libtsutil.3.dylib 0x000105af458f _ink_assert + 271
3 traffic_server 0x0001051c9ed5
_ZN19UDPReadContinuationD1Ev + 85
4 traffic_server 0x0001051ca056
_ZN14ClassAllocatorI19UDPReadContinuationE4$_44D1Ev + 24
5 traffic_server 0x0001051d452d
_ZN14ClassAllocatorI19UDPReadContinuationED1Ev + 37
6 traffic_server 0x0001051ca07b __tcf_3 + 27
7 libsystem_c.dylib 0x7fff85e547c8 __cxa_finalize + 274
8 libsystem_c.dylib 0x7fff85e54652 exit + 18
9 traffic_server 0x000104ff0b18
_ZN10UrlRewrite17load_remap_pluginEPPciP11url_mappingS0_iiPi + 4552
10 traffic_server 0x000104ffa383
_ZN10UrlRewrite10BuildTableEv + 12757
11 traffic_server 0x000104ffdcf5
_ZN10UrlRewriteC1EPKc + 2165
12 traffic_server 0x000104ef9ada
_Z18init_reverse_proxyv + 138
13 traffic_server 0x000104f623bd
_Z20init_HttpProxyServerv + 13
14 traffic_server 0x000104ecebd9 main + 6073
15 traffic_server 0x000104e582a4 start + 52
[Oct 15 14:51:28.747] Manager {0x7fff739b6960} ERROR:
[LocalManager::pollMgmtProcessServer] Server Process terminated due to Sig 6:
Abort trap: 6
{code}
do not reload bad remap.config
--
Key: TS-948
URL: https://issues.apache.org/jira/browse/TS-948
Project: Traffic Server
Issue Type: Improvement
Components: Configuration, Management
Affects Versions: 3.1.0, 3.0.1
Reporter: Conan Wang
Assignee: Leif Hedstrom
Priority: Critical
Fix For: 3.1.1
traffic_server will exit if exists bad rules in remap.config, whenever
startup or reload ( traffic_line -x ).
If remap.config is not edited correctly, the server/cluster will crash when
reloading.
It's better to pre-check the remap table and not switch to the new table if
remap.config is not enough correct.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (TS-948) do not reload bad remap.config
[ https://issues.apache.org/jira/browse/TS-948?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13128222#comment-13128222 ] Leif Hedstrom commented on TS-948: -- Argh, I missed one place ... Fixing it now, thanks for cathing it. And just to be sure, note that we'll fail to startup the first time if the remap file is broken. This is intentional, and I would be hard to convince otherwise. What the changes do is to prevent a running server to exit when you reload the config. do not reload bad remap.config -- Key: TS-948 URL: https://issues.apache.org/jira/browse/TS-948 Project: Traffic Server Issue Type: Improvement Components: Configuration, Management Affects Versions: 3.1.0, 3.0.1 Reporter: Conan Wang Assignee: Leif Hedstrom Priority: Critical Fix For: 3.1.1 traffic_server will exit if exists bad rules in remap.config, whenever startup or reload ( traffic_line -x ). If remap.config is not edited correctly, the server/cluster will crash when reloading. It's better to pre-check the remap table and not switch to the new table if remap.config is not enough correct. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (TS-948) do not reload bad remap.config
[ https://issues.apache.org/jira/browse/TS-948?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13126780#comment-13126780 ] Leif Hedstrom commented on TS-948: -- Let me know if this last commit fixes it. do not reload bad remap.config -- Key: TS-948 URL: https://issues.apache.org/jira/browse/TS-948 Project: Traffic Server Issue Type: Improvement Components: Configuration, Management Affects Versions: 3.1.0, 3.0.1 Reporter: Conan Wang Assignee: Leif Hedstrom Priority: Critical Fix For: 3.1.1 traffic_server will exit if exists bad rules in remap.config, whenever startup or reload ( traffic_line -x ). If remap.config is not edited correctly, the server/cluster will crash when reloading. It's better to pre-check the remap table and not switch to the new table if remap.config is not enough correct. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (TS-948) do not reload bad remap.config
[ https://issues.apache.org/jira/browse/TS-948?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13127257#comment-13127257 ] Conan Wang commented on TS-948: --- Fix that. But returning TS_ERROR in TSRemapInit also crashes. Sorry for missing that test in advance. do not reload bad remap.config -- Key: TS-948 URL: https://issues.apache.org/jira/browse/TS-948 Project: Traffic Server Issue Type: Improvement Components: Configuration, Management Affects Versions: 3.1.0, 3.0.1 Reporter: Conan Wang Assignee: Leif Hedstrom Priority: Critical Fix For: 3.1.1 traffic_server will exit if exists bad rules in remap.config, whenever startup or reload ( traffic_line -x ). If remap.config is not edited correctly, the server/cluster will crash when reloading. It's better to pre-check the remap table and not switch to the new table if remap.config is not enough correct. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (TS-948) do not reload bad remap.config
[ https://issues.apache.org/jira/browse/TS-948?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13125265#comment-13125265 ] Leif Hedstrom commented on TS-948: -- Please try the changes on trunk, I'm closing this for now. do not reload bad remap.config -- Key: TS-948 URL: https://issues.apache.org/jira/browse/TS-948 Project: Traffic Server Issue Type: Improvement Components: Configuration, Management Affects Versions: 3.1.0, 3.0.1 Reporter: Conan Wang Assignee: Leif Hedstrom Priority: Critical Fix For: 3.1.1 traffic_server will exit if exists bad rules in remap.config, whenever startup or reload ( traffic_line -x ). If remap.config is not edited correctly, the server/cluster will crash when reloading. It's better to pre-check the remap table and not switch to the new table if remap.config is not enough correct. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (TS-948) do not reload bad remap.config
[
https://issues.apache.org/jira/browse/TS-948?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13125568#comment-13125568
]
Conan Wang commented on TS-948:
---
My test:
1. duplicate remap rule and adding non-exist remap plugin now will not crash
running ATS. I can see WARNING: failed to reload remap.config, not replacing!
in traffic.out.
2. though bad remap.config is not reloaded, seem TSRemapDeleteInstance
(TSfree ihandle) of my remap plugin is called according to my debug log.
However, my remap plugin work well like before, that is, I can still retrieve
the stored variable in ihandle. I don't know whether there is potential problem.
3. return TS_ERROR in TSRemapNewInstance still crash.
{code}
Server {0x1079f6000} ERROR: Failed to create new instance for plugin
/Users/conan/box/ts-trunk/libexec/trafficserver/add_header.so (non-zero
retval)... bailing out
FATAL: UnixUDPNet.cc:295: failed assert `event != NULL`
/Users/conan/box/ts-trunk/bin/traffic_server - STACK TRACE:
0 libtsutil.3.dylib 0x000102fb05ab ink_fatal_va + 283
1 libtsutil.3.dylib 0x000102fb08b4 ink_fatal + 356
2 libtsutil.3.dylib 0x000102fadcff _ink_assert + 271
3 traffic_server 0x000102684f45
_ZN19UDPReadContinuationD1Ev + 85
4 traffic_server 0x0001026850c6
_ZN14ClassAllocatorI19UDPReadContinuationE4$_44D1Ev + 24
5 traffic_server 0x00010268f59d
_ZN14ClassAllocatorI19UDPReadContinuationED1Ev + 37
6 traffic_server 0x0001026850eb __tcf_3 + 27
7 libsystem_c.dylib 0x7fff872217c8 __cxa_finalize + 274
8 libsystem_c.dylib 0x7fff87221652 exit + 18
9 traffic_server 0x0001024aca9c
_ZN10UrlRewrite17load_remap_pluginEPPciP11url_mappingS0_iiPi + 8284
10 traffic_server 0x0001024b5473
_ZN10UrlRewrite10BuildTableEv + 12757
11 traffic_server 0x0001024b8dc5
_ZN10UrlRewriteC1EPKc + 2165
12 traffic_server 0x0001023b450e
_Z16reloadUrlRewritev + 350
13 traffic_server 0x0001023b5296
_ZN21UR_UpdateContinuation19file_update_handlerEiPv + 24
14 traffic_server 0x0001023d7434
_ZN12Continuation11handleEventEiPv + 148
15 traffic_server 0x0001026a787a
_ZN7EThread13process_eventEP5Eventi + 418
16 traffic_server 0x0001026a6c85
_ZN7EThread7executeEv + 191
17 traffic_server 0x0001026a6174
_ZL21spawn_thread_internalPv + 132
18 libsystem_c.dylib 0x7fff8722e8bf _pthread_start + 335
19 libsystem_c.dylib 0x7fff87231b75 thread_start + 13
{code}
do not reload bad remap.config
--
Key: TS-948
URL: https://issues.apache.org/jira/browse/TS-948
Project: Traffic Server
Issue Type: Improvement
Components: Configuration, Management
Affects Versions: 3.1.0, 3.0.1
Reporter: Conan Wang
Assignee: Leif Hedstrom
Priority: Critical
Fix For: 3.1.1
traffic_server will exit if exists bad rules in remap.config, whenever
startup or reload ( traffic_line -x ).
If remap.config is not edited correctly, the server/cluster will crash when
reloading.
It's better to pre-check the remap table and not switch to the new table if
remap.config is not enough correct.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (TS-948) do not reload bad remap.config
[ https://issues.apache.org/jira/browse/TS-948?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13125573#comment-13125573 ] Leif Hedstrom commented on TS-948: -- On 2, yeas, as expected. It still goes through all instantiatiobs, but unwinds on failure . Since this is no different for your plugin whether it succeeds or not, it shouldn't matter. You still have to deal with proper uninstantiation on every reload. On 3, bah, missed that special case, nice catch. I'll look at it tomorrow. do not reload bad remap.config -- Key: TS-948 URL: https://issues.apache.org/jira/browse/TS-948 Project: Traffic Server Issue Type: Improvement Components: Configuration, Management Affects Versions: 3.1.0, 3.0.1 Reporter: Conan Wang Assignee: Leif Hedstrom Priority: Critical Fix For: 3.1.1 traffic_server will exit if exists bad rules in remap.config, whenever startup or reload ( traffic_line -x ). If remap.config is not edited correctly, the server/cluster will crash when reloading. It's better to pre-check the remap table and not switch to the new table if remap.config is not enough correct. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa For more information on JIRA, see: http://www.atlassian.com/software/jira
