Re: [Apache Bloodhound] #613: Add / remove milestone breaks a whole product
#613: Add / remove milestone breaks a whole product
---+---
Reporter: olemis| Owner: nobody
Type: defect| Status: new
Priority: blocker | Milestone: next 0.x
Component: multiproduct |Version: 0.6.0
Resolution:| Keywords: milestone, postgresql, db, sqlite
---+---
Comment (by wimh):
I currently can't respond to the lru_cache decorator, I did not look into
that part of the code.
But I think there are two separate issues with the adding/removing of
milestones.
= !QuickCreateTicketDialog =
If I disable the !QuickCreateTicketDialog from the !BloodhoundTheme
plugin, this error will not be shown again after deleting a milestone:
{{{
Error: Invalid milestone name
Milestone milestone2 does not exist.
}}}
= !MilestoneCache =
But still after adding or deleting milestones, the changes are not always
visible in the list. If I remove the {{{@cached}}} decorator from
source:/trunk/trac/trac/ticket/model.py#L949 (ie replace it with
{{{@property}}} and remove both lines containing {{{del
self.cache.milestones}}}), that is fixed too. But this alone is not
sufficient to fix the !QuickCreateTicketDialog.
The first issue seems to resolve itself after a few minutes. I am not sure
about the second issue, it looks like apache must be restarted to fix
that.
Note that another easy way to reproduce this is running two {{{tracd}}}
processes on different ports.
--
Ticket URL: https://issues.apache.org/bloodhound/ticket/613#comment:19
Apache Bloodhound https://issues.apache.org/bloodhound/
The Apache Bloodhound issue tracker
svn commit: r1611921 - in /bloodhound/trunk/bloodhound_multiproduct/multiproduct: env.py hooks.py
Author: gjm
Date: Sat Jul 19 17:07:48 2014
New Revision: 1611921
URL: http://svn.apache.org/r1611921
Log:
clearing the lru_cache close to the beginning of a request to limit impact of
caching errors in multi process/threaded web serving - towards #613
Modified:
bloodhound/trunk/bloodhound_multiproduct/multiproduct/env.py
bloodhound/trunk/bloodhound_multiproduct/multiproduct/hooks.py
Modified: bloodhound/trunk/bloodhound_multiproduct/multiproduct/env.py
URL:
http://svn.apache.org/viewvc/bloodhound/trunk/bloodhound_multiproduct/multiproduct/env.py?rev=1611921r1=1611920r2=1611921view=diff
==
--- bloodhound/trunk/bloodhound_multiproduct/multiproduct/env.py (original)
+++ bloodhound/trunk/bloodhound_multiproduct/multiproduct/env.py Sat Jul 19
17:07:48 2014
@@ -365,6 +365,8 @@ class ProductEnvironment(Component, Comp
g_env = env.parent if isinstance(env,
ProductEnvironment) else env
return f(self, g_env, *args, **kwargs)
+__call__.clear = f.clear
+
return __call__
def product_env_keymap(args, kwds, kwd_mark):
@@ -397,6 +399,10 @@ class ProductEnvironment(Component, Comp
setup_participants = ExtensionPoint(trac.env.IEnvironmentSetupParticipant)
multi_product_support_components =
ExtensionPoint(ISupportMultiProductEnvironment)
+@classmethod
+def clear_env_cache(cls):
+cls.__metaclass__.__call__.clear()
+
@property
def product_setup_participants(self):
return [
Modified: bloodhound/trunk/bloodhound_multiproduct/multiproduct/hooks.py
URL:
http://svn.apache.org/viewvc/bloodhound/trunk/bloodhound_multiproduct/multiproduct/hooks.py?rev=1611921r1=1611920r2=1611921view=diff
==
--- bloodhound/trunk/bloodhound_multiproduct/multiproduct/hooks.py (original)
+++ bloodhound/trunk/bloodhound_multiproduct/multiproduct/hooks.py Sat Jul 19
17:07:48 2014
@@ -37,6 +37,8 @@ PRODUCT_RE = re.compile(r'^/products(?:/
class MultiProductEnvironmentFactory(EnvironmentFactoryBase):
def open_environment(self, environ, env_path, global_env, use_cache=False):
+# clearing product environment cache - bh:ticket:613
+multiproduct.env.ProductEnvironment.clear_env_cache()
environ.setdefault('SCRIPT_NAME', '') # bh:ticket:594
env = pid = product_path = None
Re: [Apache Bloodhound] #613: Add / remove milestone breaks a whole product
#613: Add / remove milestone breaks a whole product
---+---
Reporter: olemis| Owner: gjm
Type: defect| Status: accepted
Priority: blocker | Milestone: next 0.x
Component: multiproduct |Version: 0.6.0
Resolution:| Keywords: milestone, postgresql, db, sqlite
---+---
Changes (by gjm):
* owner: nobody = gjm
* status: new = accepted
Comment:
patch from comment:18 committed in r1611921
I think wimh's findings are still consistent with my analysis and there
are multiple symptoms resulting from the same underlying problem. Running
multiple tracd processes can be considered roughly the same as having
{{{processes=2}}} for the {{{WSGIDaemonProcess}}}.
To be fair I haven't replicated any errors around this that I believe
resolve themselves after a few minutes as the errors I observed will
continue to appear intermittently until the webserver restarts.
With luck the commit is a good enough solution for now.
--
Ticket URL: https://issues.apache.org/bloodhound/ticket/613#comment:20
Apache Bloodhound https://issues.apache.org/bloodhound/
The Apache Bloodhound issue tracker
[Apache Bloodhound] BloodhoundInstall modified
Page BloodhoundInstall was changed by wimh
Diff URL:
https://issues.apache.org/bloodhound/wiki/BloodhoundInstall?action=diffversion=30
Revision 30
Comment: additional instructions for PostgreSQL
Changes:
---8--8--8--8--8--8--8--8
Index: BloodhoundInstall
=
--- BloodhoundInstall (version: 29)
+++ BloodhoundInstall (version: 30)
@@ -104,6 +104,12 @@
source ./bloodhound/bin/activate
pip install -r requirements.txt
}}}
+
+If you are using PostgreSQL, you also need to run:
+ {{{#!sh
+ pip install -r pgrequirements.txt
+ }}}
+If this fails, you might need you to install the python-dev and postgres-dev
packages and try again.
Then run the {{{bloodhound_setup.py}}} script:
{{{#!sh
---8--8--8--8--8--8--8--8
--
Page URL: https://issues.apache.org/bloodhound/wiki/BloodhoundInstall
Apache Bloodhound https://issues.apache.org/bloodhound/
The Apache Bloodhound issue tracker
This is an automated message. Someone added your email address to be
notified of changes on 'BloodhoundInstall' page.
If it was not you, please report to .
Re: [Apache Bloodhound] #613: Add / remove milestone breaks a whole product
#613: Add / remove milestone breaks a whole product
---+---
Reporter: olemis| Owner: gjm
Type: defect| Status: accepted
Priority: blocker | Milestone: next 0.x
Component: multiproduct |Version: 0.6.0
Resolution:| Keywords: milestone, postgresql, db, sqlite
---+---
Comment (by wimh):
Replying to [comment:20 gjm]:
With luck the commit is a good enough solution for now.
It fixes the problems for me, but do I understand it correctly it flushes
all caches each time for each request? At least in the
{{{MilestoneCache}}} it looks like a new SQL query is done each time I
refresh the milestones admin page.
--
Ticket URL: https://issues.apache.org/bloodhound/ticket/613#comment:21
Apache Bloodhound https://issues.apache.org/bloodhound/
The Apache Bloodhound issue tracker
