[
https://issues.apache.org/jira/browse/TS-4925?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15592517#comment-15592517
]
Shrihari commented on TS-4925:
------------------------------
Guys,
We hit this issue as well and our RCA was a bit different. We found that it was
a regression of https://issues.apache.org/jira/browse/TS-3960. In that fix,
this change was added:
+// void configFileChild(const char *parent, const char *child)
+//
+// Add child to the bindings with parentRollback
+void
+FileManager::configFileChild(const char *parent, const char *child, unsigned
int options)
+{
+ InkHashTableValue lookup;
+ Rollback *parentRollback = NULL;
+ ink_mutex_acquire(&accessLock);
+ int htfound = ink_hash_table_lookup(bindings, parent, &lookup);
+ if (htfound) {
+ parentRollback = (Rollback *)lookup;
+ }
+ RollBackVersionType version_type = options ? ROLLBACK_VERSIONED :
ROLLBACK_UNVERSIONED;
+ if (htfound) {
+ addFileHelper(child, true, parentRollback, version_type);
+ }
+ ink_mutex_release(&accessLock);
+}
+
Which means that all the files added using this permission have
'root_access_needed' set to true. Currently only ssl_certificates use this
function because of the relationship to ssl_multicert.
We realized that one of the traffic_manager threads ('sync_thr' to be precise)
keeps attempting to 'elevate' the privileges. This results in receiving SIG_RT
(signal 33) which interrupts the syscall select every few seconds.
We fixed this by passing 'root_access_needed' flag from traffic_server instead
of hardcoding it to 'true', we never saw these errors again. If this logic is
acceptable, please let me know I will try to submit a patch for this.
Thanks.
> Manager puking EBADF everywhere
> -------------------------------
>
> Key: TS-4925
> URL: https://issues.apache.org/jira/browse/TS-4925
> Project: Traffic Server
> Issue Type: Bug
> Components: Manager
> Reporter: James Peach
> Assignee: James Peach
> Fix For: 7.0.0
>
> Time Spent: 1h 10m
> Remaining Estimate: 0h
>
> {noformat}
> [Oct 3 20:19:53.824] Manager {0x7f7b699a2800} ERROR:
> [LocalManager::pollMgmtProcessServer] select failed or was interrupted (9)
> ...
> {noformat}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
