Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

t2008-althash.t errors #1063

Closed
chu11 opened this issue May 16, 2017 · 4 comments
Closed

t2008-althash.t errors #1063

chu11 opened this issue May 16, 2017 · 4 comments

Comments

@chu11
Copy link
Member

chu11 commented May 16, 2017

I'm seeing on catalyst and hype2 this sometimes

not ok 5 - Attempt to start instance with invalid hash fails hard

with debug

lt-flux-broker: configuring attributes: Invalid argument
E: (flux-broker) 17-05-16 15:10:03 dangling 'PAIR' socket created at src/zsys.c:379
E: (flux-broker) 17-05-16 15:10:03 dangling 'PAIR' socket created at src/zsys.c:380
E: (flux-broker) 17-05-16 15:10:03 dangling 'REP' socket created at src/zauth.c:77
E: (flux-broker) 17-05-16 15:10:03 dangling 'ROUTER' socket created at overlay.c:479
E: (flux-broker) 17-05-16 15:10:03 dangling sockets: cannot terminate ZMQ safely
lt-flux-broker: src/zsock_option.c:1347: zsock_set_sndtimeo: Assertion `rc == 0 || zmq_errno () == (156384712 + 53)' failed.
./sharness.sh: line 504: 27842 Aborted                 (core dumped) "$@"
test_must_fail: died by signal: flux start -o,-Scontent.hash=wronghash /bin/true
not ok 5 - Attempt to start instance with invalid hash fails hard
#	
#	    test_must_fail flux start -o,-Scontent.hash=wronghash /bin/true
#	

It appears #1036 causes a failure that's "too hard" of a failure

@garlick
Copy link
Member

garlick commented May 16, 2017

FWIW: Just hit this one in c9.io (Ubuntu 14.04.2 LTS), though in a dozen manual runs I couldn't reproduce it.

@chu11
Copy link
Member Author

chu11 commented May 17, 2017

For me on catalyst it is occurring quite frequently, perhaps atleast 50% of the time.

@garlick
Copy link
Member

garlick commented May 17, 2017

Looks like content_cache_register_attrs () could be called much earlier. I wonder if that helps?

diff --git a/src/broker/broker.c b/src/broker/broker.c
index 199f1cd..a33c939 100644
--- a/src/broker/broker.c
+++ b/src/broker/broker.c
@@ -348,6 +348,8 @@ int main (int argc, char *argv[])
         if ((e = argz_create (argv + optind, &ctx.init_shell_cmd, &ctx.init_shell_cmd_len)) != 0)
             log_errn_exit (e, "argz_create");
     }
+    if (content_cache_register_attrs (ctx.cache, ctx.attrs) < 0)
+        log_err_exit ("content cache attributes");
 
     /* Record the instance owner: the effective uid of the broker.
      * Set default rolemask for messages sent with flux_send()
@@ -497,8 +499,7 @@ int main (int argc, char *argv[])
             || attr_add_active_int (ctx.attrs, "tbon.descendants",
                                 &ctx.tbon.descendants,
                                 FLUX_ATTRFLAG_IMMUTABLE) < 0
-            || hello_register_attrs (ctx.hello, ctx.attrs) < 0
-            || content_cache_register_attrs (ctx.cache, ctx.attrs) < 0) {
+            || hello_register_attrs (ctx.hello, ctx.attrs) < 0) {
         log_err_exit ("configuring attributes");
     }

@chu11
Copy link
Member Author

chu11 commented May 17, 2017

It does seem to help. make -j16 check passed 8/8 times, which was much better success than what I had before.

@grondo grondo closed this as completed in 3d59fbf May 28, 2017
chu11 added a commit to chu11/flux-core that referenced this issue Jun 15, 2019
Move all "register_attrs" calls to a location closer to the
initialization / setup of other parts of the object.  With this
change, the register_attrs calls are more in line with their
object and removes the impression that the registering of the
attrs needs to be done in a specific order except for when
commented.

The movement of content_cache_register_attrs() effectively
reverses commit

3d59fbf

which was a fix for flux-framework#1063.  The fix of issue flux-framework#1036 no longer
requires the content_cache_register_attrs() call to be made
far earlier in the code before zsys_init().
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants