You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@qpid.apache.org by Alan Conway <ac...@redhat.com> on 2013/04/09 17:57:07 UTC
Review Request: HA migration guide
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/10376/
-----------------------------------------------------------
Review request for qpid, Andy Goldstein, Jason Dillaman, Pavel Moravec, and Justin Ross.
Description
-------
This is the first draft of HA migration guide, I'm hoping to get some review feedback.
The patch also has some minor updates to the HA chapter of the Qpid broker book, they're less important for review.
Diffs
-----
/branches/0.22/qpid/cpp/README-HA.txt PRE-CREATION
/branches/0.22/qpid/doc/book/src/cpp-broker/Active-Passive-Cluster.xml 1464646
Diff: https://reviews.apache.org/r/10376/diff/
Testing
-------
Thanks,
Alan Conway
Re: Review Request: NO-JIRA: HA updates to migration guide.
Posted by Alan Conway <ac...@redhat.com>.
> On April 11, 2013, 4:44 p.m., Jason Dillaman wrote:
> > /trunk/qpid/doc/book/src/cpp-broker/Active-Passive-Cluster.xml, line 180
> > <https://reviews.apache.org/r/10376/diff/2/?file=280440#file280440line180>
> >
> > Might want to clarify that only outbound links will be lost -- any inbound federated connections will persist and recover. I know this confused one end user when they read this and thought federation would break after a failover. Also, might be helpful to point out that the 'qpid-primary'-like init script can be modified to re-add any static federated links upon promotion to primary.
Federation links <emphasis>to</emphasis> the primary will fail over
correctly. Federated links <emphasis>from</emphasis> the primary
will be lost in fail over, they will not be re-connected to the new
primary. It is possible to work around this by replacing the
<literal>qpidd-primary</literal> start up script with a script that
re-creates federation links when the primary is promoted.
- Alan
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/10376/#review19011
-----------------------------------------------------------
On April 11, 2013, 4:04 p.m., Alan Conway wrote:
>
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/10376/
> -----------------------------------------------------------
>
> (Updated April 11, 2013, 4:04 p.m.)
>
>
> Review request for qpid, Andy Goldstein, Jason Dillaman, mick goulish, Pavel Moravec, and Justin Ross.
>
>
> Description
> -------
>
> NO-JIRA: HA updates to migration guide.
>
> git-svn-id: https://svn.apache.org/repos/asf/qpid/trunk@1466933 13f79535-47bb-0310-9956-ffa450edef68
>
>
> Diffs
> -----
>
> /trunk/qpid/cpp/README-HA.txt 1466895
> /trunk/qpid/doc/book/src/cpp-broker/Active-Passive-Cluster.xml 1466895
>
> Diff: https://reviews.apache.org/r/10376/diff/
>
>
> Testing
> -------
>
>
> Thanks,
>
> Alan Conway
>
>
Re: Review Request: NO-JIRA: HA updates to migration guide.
Posted by Jason Dillaman <jd...@redhat.com>.
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/10376/#review19011
-----------------------------------------------------------
/trunk/qpid/doc/book/src/cpp-broker/Active-Passive-Cluster.xml
<https://reviews.apache.org/r/10376/#comment39500>
Might want to clarify that only outbound links will be lost -- any inbound federated connections will persist and recover. I know this confused one end user when they read this and thought federation would break after a failover. Also, might be helpful to point out that the 'qpid-primary'-like init script can be modified to re-add any static federated links upon promotion to primary.
- Jason Dillaman
On April 11, 2013, 4:04 p.m., Alan Conway wrote:
>
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/10376/
> -----------------------------------------------------------
>
> (Updated April 11, 2013, 4:04 p.m.)
>
>
> Review request for qpid, Andy Goldstein, Jason Dillaman, mick goulish, Pavel Moravec, and Justin Ross.
>
>
> Description
> -------
>
> NO-JIRA: HA updates to migration guide.
>
> git-svn-id: https://svn.apache.org/repos/asf/qpid/trunk@1466933 13f79535-47bb-0310-9956-ffa450edef68
>
>
> Diffs
> -----
>
> /trunk/qpid/cpp/README-HA.txt 1466895
> /trunk/qpid/doc/book/src/cpp-broker/Active-Passive-Cluster.xml 1466895
>
> Diff: https://reviews.apache.org/r/10376/diff/
>
>
> Testing
> -------
>
>
> Thanks,
>
> Alan Conway
>
>
Re: Review Request: NO-JIRA: HA updates to migration guide.
Posted by Pavel Moravec <pm...@redhat.com>.
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/10376/#review19856
-----------------------------------------------------------
My comments (sorry for later review):
- it is not stated what ha-* broker options are mandatory and what are optional.
- `link-*-interval` options are valid for inter-broker links, I guess (but not evident from the description). I suggest changing:
- "to check link health" to "to check inter-broker management/failover/whatever-fits-best link health", and
- "for replication links" to "for inter-broker replication links"
(both in README-HA.txt and the xml file)
- cluster.conf example has VIP 20.0.20.200 while further 20.0.10.200 is used (e.g. ha-public-url and in text)
- MECH vs. MECHANISM: compare lines 302 and 797 of the xml
- Pavel Moravec
On April 11, 2013, 4:04 p.m., Alan Conway wrote:
>
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/10376/
> -----------------------------------------------------------
>
> (Updated April 11, 2013, 4:04 p.m.)
>
>
> Review request for qpid, Andy Goldstein, Jason Dillaman, mick goulish, Pavel Moravec, and Justin Ross.
>
>
> Description
> -------
>
> NO-JIRA: HA updates to migration guide.
>
> git-svn-id: https://svn.apache.org/repos/asf/qpid/trunk@1466933 13f79535-47bb-0310-9956-ffa450edef68
>
>
> Diffs
> -----
>
> /trunk/qpid/cpp/README-HA.txt 1466895
> /trunk/qpid/doc/book/src/cpp-broker/Active-Passive-Cluster.xml 1466895
>
> Diff: https://reviews.apache.org/r/10376/diff/
>
>
> Testing
> -------
>
>
> Thanks,
>
> Alan Conway
>
>
Re: Review Request: NO-JIRA: HA updates to migration guide.
Posted by Jason Dillaman <jd...@redhat.com>.
> On April 11, 2013, 4:34 p.m., Jason Dillaman wrote:
> > /trunk/qpid/doc/book/src/cpp-broker/Active-Passive-Cluster.xml, line 269
> > <https://reviews.apache.org/r/10376/diff/2/?file=280440#file280440line269>
> >
> > New behavior supports 'none' keyword and does not default to the broker url
>
> Alan Conway wrote:
> What was the use-case for "none" again?
It's original intention was to match the behavior of the 'known-hosts-url' parameter to indicate that the URL should not be defaulted and instead be left blank. However, given that the HA plugin no longer defaults the URL when the parameter is not provided, I think it is no longer needed.
- Jason
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/10376/#review19008
-----------------------------------------------------------
On April 11, 2013, 4:04 p.m., Alan Conway wrote:
>
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/10376/
> -----------------------------------------------------------
>
> (Updated April 11, 2013, 4:04 p.m.)
>
>
> Review request for qpid, Andy Goldstein, Jason Dillaman, mick goulish, Pavel Moravec, and Justin Ross.
>
>
> Description
> -------
>
> NO-JIRA: HA updates to migration guide.
>
> git-svn-id: https://svn.apache.org/repos/asf/qpid/trunk@1466933 13f79535-47bb-0310-9956-ffa450edef68
>
>
> Diffs
> -----
>
> /trunk/qpid/cpp/README-HA.txt 1466895
> /trunk/qpid/doc/book/src/cpp-broker/Active-Passive-Cluster.xml 1466895
>
> Diff: https://reviews.apache.org/r/10376/diff/
>
>
> Testing
> -------
>
>
> Thanks,
>
> Alan Conway
>
>
Re: Review Request: NO-JIRA: HA updates to migration guide.
Posted by Alan Conway <ac...@redhat.com>.
> On April 11, 2013, 4:34 p.m., Jason Dillaman wrote:
> > /trunk/qpid/cpp/README-HA.txt, line 79
> > <https://reviews.apache.org/r/10376/diff/2/?file=280439#file280439line79>
> >
> > I believe rgmanager is deprecated in upstream Fedora. Might want to indicate that using rgmanager is just one option and not a requirement.
The new HA module requires an external cluster resource manager. Initially it
supports `rgmanager` provided by the `cman` package. `rgmanager` is responsible
for stopping and starting brokers and determining which broker is promoted to
primary. It is also responsible for detecting primary failures and promoting a
new primary. See ["Configuring rgmanager as resource manager"][ha-rm-config]
It is relatively easy to integrate with a new resource manager, see
["Integrating with other Cluster Resource Managers"][ha-other-rm]
> On April 11, 2013, 4:34 p.m., Jason Dillaman wrote:
> > /trunk/qpid/doc/book/src/cpp-broker/Active-Passive-Cluster.xml, line 58
> > <https://reviews.apache.org/r/10376/diff/2/?file=280440#file280440line58>
> >
> > -- or delays the ack until the the received message has been consumed from its destination queues.
Good point:
<title>Avoiding message loss</title>
<para>
In order to avoid message loss, the primary broker <emphasis>delays
acknowledgment</emphasis> of messages received from clients until the
message has been replicated and acknowledged by all of the back-up
brokers, or has been consumed from the primary queue.
</para>
<para>
This ensures that all acknowledged messages are safe: they have either
been consumed or backed up to all backup brokers. Messages that are
consumed <emphasis>before</emphasis> they are replicated do not need to
be replicated. This reduces the work load when replicating a queue with
active consumers.
</para>
> On April 11, 2013, 4:34 p.m., Jason Dillaman wrote:
> > /trunk/qpid/doc/book/src/cpp-broker/Active-Passive-Cluster.xml, line 164
> > <https://reviews.apache.org/r/10376/diff/2/?file=280440#file280440line164>
> >
> > Not a documentation issue, but should this limitation be fixed prior to removing support for active-active clustering?
Yes, it's on the roadmap.
> On April 11, 2013, 4:34 p.m., Jason Dillaman wrote:
> > /trunk/qpid/doc/book/src/cpp-broker/Active-Passive-Cluster.xml, line 269
> > <https://reviews.apache.org/r/10376/diff/2/?file=280440#file280440line269>
> >
> > New behavior supports 'none' keyword and does not default to the broker url
What was the use-case for "none" again?
- Alan
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/10376/#review19008
-----------------------------------------------------------
On April 11, 2013, 4:04 p.m., Alan Conway wrote:
>
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/10376/
> -----------------------------------------------------------
>
> (Updated April 11, 2013, 4:04 p.m.)
>
>
> Review request for qpid, Andy Goldstein, Jason Dillaman, mick goulish, Pavel Moravec, and Justin Ross.
>
>
> Description
> -------
>
> NO-JIRA: HA updates to migration guide.
>
> git-svn-id: https://svn.apache.org/repos/asf/qpid/trunk@1466933 13f79535-47bb-0310-9956-ffa450edef68
>
>
> Diffs
> -----
>
> /trunk/qpid/cpp/README-HA.txt 1466895
> /trunk/qpid/doc/book/src/cpp-broker/Active-Passive-Cluster.xml 1466895
>
> Diff: https://reviews.apache.org/r/10376/diff/
>
>
> Testing
> -------
>
>
> Thanks,
>
> Alan Conway
>
>
Re: Review Request: NO-JIRA: HA updates to migration guide.
Posted by Jason Dillaman <jd...@redhat.com>.
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/10376/#review19008
-----------------------------------------------------------
/trunk/qpid/cpp/README-HA.txt
<https://reviews.apache.org/r/10376/#comment39488>
I believe rgmanager is deprecated in upstream Fedora. Might want to indicate that using rgmanager is just one option and not a requirement.
/trunk/qpid/doc/book/src/cpp-broker/Active-Passive-Cluster.xml
<https://reviews.apache.org/r/10376/#comment39490>
-- or delays the ack until the the received message has been consumed from its destination queues.
/trunk/qpid/doc/book/src/cpp-broker/Active-Passive-Cluster.xml
<https://reviews.apache.org/r/10376/#comment39492>
"If the primary crashes" instead of "So if"
/trunk/qpid/doc/book/src/cpp-broker/Active-Passive-Cluster.xml
<https://reviews.apache.org/r/10376/#comment39493>
Not a documentation issue, but should this limitation be fixed prior to removing support for active-active clustering?
/trunk/qpid/doc/book/src/cpp-broker/Active-Passive-Cluster.xml
<https://reviews.apache.org/r/10376/#comment39495>
New behavior supports 'none' keyword and does not default to the broker url
- Jason Dillaman
On April 11, 2013, 4:04 p.m., Alan Conway wrote:
>
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/10376/
> -----------------------------------------------------------
>
> (Updated April 11, 2013, 4:04 p.m.)
>
>
> Review request for qpid, Andy Goldstein, Jason Dillaman, mick goulish, Pavel Moravec, and Justin Ross.
>
>
> Description
> -------
>
> NO-JIRA: HA updates to migration guide.
>
> git-svn-id: https://svn.apache.org/repos/asf/qpid/trunk@1466933 13f79535-47bb-0310-9956-ffa450edef68
>
>
> Diffs
> -----
>
> /trunk/qpid/cpp/README-HA.txt 1466895
> /trunk/qpid/doc/book/src/cpp-broker/Active-Passive-Cluster.xml 1466895
>
> Diff: https://reviews.apache.org/r/10376/diff/
>
>
> Testing
> -------
>
>
> Thanks,
>
> Alan Conway
>
>
Re: Review Request: NO-JIRA: HA updates to migration guide.
Posted by Alan Conway <ac...@redhat.com>.
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/10376/
-----------------------------------------------------------
(Updated April 11, 2013, 4:04 p.m.)
Review request for qpid, Andy Goldstein, Jason Dillaman, mick goulish, Pavel Moravec, and Justin Ross.
Changes
-------
Updated with Andy's feedback.
Summary (updated)
-----------------
NO-JIRA: HA updates to migration guide.
Description (updated)
-------
NO-JIRA: HA updates to migration guide.
git-svn-id: https://svn.apache.org/repos/asf/qpid/trunk@1466933 13f79535-47bb-0310-9956-ffa450edef68
Diffs (updated)
-----
/trunk/qpid/cpp/README-HA.txt 1466895
/trunk/qpid/doc/book/src/cpp-broker/Active-Passive-Cluster.xml 1466895
Diff: https://reviews.apache.org/r/10376/diff/
Testing
-------
Thanks,
Alan Conway
Re: Review Request: HA migration guide
Posted by Alan Conway <ac...@redhat.com>.
> On April 9, 2013, 4:06 p.m., Andy Goldstein wrote:
> > /branches/0.22/qpid/cpp/README-HA.txt, line 23
> > <https://reviews.apache.org/r/10376/diff/1/?file=279331#file279331line23>
> >
> > a client will only fail over to other URLs if the connection options include multiple URLs - if the client only specifies 1 URL, and it happens to point to a broker that is currently the backup, it will never connect... right?
It can be a virtual IP, clarified text:
`ha-brokers-url`: A URL listing each node in the cluster. For example
`amqp:node1.example.com,node2.example.com,node3.example.com`. If you have separate
client and broker networks, the addresses in `ha-brokers-url` should be on the
broker network. Note this URL must list all nodes in the cluster, you cannot use
a [Virtual IP address][ha-virtual-ip] here.
`ha-public-url`: URL used by clients to connect to the cluster.
This can be one of:
- A [virtual IP address][ha-virtual-ip] configured for your client network.
- A list of broker addresses on the client network. If you don't have a separate
client network then this is the same as the `ha-brokers-url`
> On April 9, 2013, 4:06 p.m., Andy Goldstein wrote:
> > /branches/0.22/qpid/doc/book/src/cpp-broker/Active-Passive-Cluster.xml, line 613
> > <https://reviews.apache.org/r/10376/diff/1/?file=279332#file279332line613>
> >
> > assuming the client is configured with a list of urls
Can be virtual IP
Clients can only connect to the primary broker. Backup brokers reject any
connection attempt by a client. Clients rejected by a backup broker will
automatically fail-over until they connect to the primary. if
<literal>ha-public-url</literal> contains multiple addresses, the client
will them all in rotation. If it is a virtual IP address the clients will
retry on the same address until it is reconnected.
- Alan
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/10376/#review18853
-----------------------------------------------------------
On April 9, 2013, 3:57 p.m., Alan Conway wrote:
>
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/10376/
> -----------------------------------------------------------
>
> (Updated April 9, 2013, 3:57 p.m.)
>
>
> Review request for qpid, Andy Goldstein, Jason Dillaman, Pavel Moravec, and Justin Ross.
>
>
> Description
> -------
>
> This is the first draft of HA migration guide, I'm hoping to get some review feedback.
>
> The patch also has some minor updates to the HA chapter of the Qpid broker book, they're less important for review.
>
>
> Diffs
> -----
>
> /branches/0.22/qpid/cpp/README-HA.txt PRE-CREATION
> /branches/0.22/qpid/doc/book/src/cpp-broker/Active-Passive-Cluster.xml 1464646
>
> Diff: https://reviews.apache.org/r/10376/diff/
>
>
> Testing
> -------
>
>
> Thanks,
>
> Alan Conway
>
>
Re: Review Request: HA migration guide
Posted by Andy Goldstein <ag...@redhat.com>.
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/10376/#review18853
-----------------------------------------------------------
/branches/0.22/qpid/cpp/README-HA.txt
<https://reviews.apache.org/r/10376/#comment39304>
a client will only fail over to other URLs if the connection options include multiple URLs - if the client only specifies 1 URL, and it happens to point to a broker that is currently the backup, it will never connect... right?
/branches/0.22/qpid/cpp/README-HA.txt
<https://reviews.apache.org/r/10376/#comment39305>
typo: exaple -> example
/branches/0.22/qpid/cpp/README-HA.txt
<https://reviews.apache.org/r/10376/#comment39306>
or other unknown behaviors
/branches/0.22/qpid/doc/book/src/cpp-broker/Active-Passive-Cluster.xml
<https://reviews.apache.org/r/10376/#comment39308>
assuming the client is configured with a list of urls
/branches/0.22/qpid/doc/book/src/cpp-broker/Active-Passive-Cluster.xml
<https://reviews.apache.org/r/10376/#comment39309>
extra space in between a and cluster
- Andy Goldstein
On April 9, 2013, 3:57 p.m., Alan Conway wrote:
>
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/10376/
> -----------------------------------------------------------
>
> (Updated April 9, 2013, 3:57 p.m.)
>
>
> Review request for qpid, Andy Goldstein, Jason Dillaman, Pavel Moravec, and Justin Ross.
>
>
> Description
> -------
>
> This is the first draft of HA migration guide, I'm hoping to get some review feedback.
>
> The patch also has some minor updates to the HA chapter of the Qpid broker book, they're less important for review.
>
>
> Diffs
> -----
>
> /branches/0.22/qpid/cpp/README-HA.txt PRE-CREATION
> /branches/0.22/qpid/doc/book/src/cpp-broker/Active-Passive-Cluster.xml 1464646
>
> Diff: https://reviews.apache.org/r/10376/diff/
>
>
> Testing
> -------
>
>
> Thanks,
>
> Alan Conway
>
>