ambari-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Matt <mithm...@gmail.com>
Subject Re: Review Request 43753: Update description for HAWQ Sync standby alert
Date Tue, 23 Feb 2016 06:15:42 GMT


> On Feb. 18, 2016, 7:57 p.m., bhuvnesh chaudhary wrote:
> > ambari-server/src/main/resources/common-services/HAWQ/2.0.0/alerts.json, line 7
> > <https://reviews.apache.org/r/43753/diff/1/?file=1258950#file1258950line7>
> >
> >     We might need to still rephrase it and may be what we had earlier was okay,
as there are situation where the status will be mentioned as Not Synchronized but resynchronize
will not fix it. Re-sync will not work on all the situations.
> >     
> >     Once has to remove the standby and re-init it. 
> >     
> >     Refer to the greenplum article, HAWQ should be behaving the same. 
> >     https://support.pivotal.io/hc/en-us/articles/201157218-gpstate-f-reports-Standby-master-too-far-behind-
> 
> Matt wrote:
>     So what are you suggesting?
>     - Leave the message as it is, and let the user figure it out somehow? Not a good
idea to leave the user in the dark.
>     - Add an additional line saying: "If HAWQ Standby is too far behind HAWQ Master,
HAWQ Standby might have to be deleted and installed again"?
> 
> Goutam Tadi wrote:
>     Suggestion to addendum: "Try resynchronizing Standby (Re-Synchronize HAWQ Standby).
If this doesn't work  reinstall Standby."
> 
> bhuvnesh chaudhary wrote:
>     It is a description of the alert, so can be informative. So will suggest to have
additional information.
>     However, before adding that line, lets syncup in Apache HAWQ forum to confirm the
behavior and then update here.
> 
> Matt wrote:
>     How about this?
>     
>     "This alert will trigger if HAWQ Standby is not synchronized with HAWQ Master. To
synchronize HAWQ Standby with HAWQ Master, use the service action 'Re-Synchronize HAWQ Standby'
on HAWQ service page. If HAWQ Standby is too far behind HAWQ Master, HAWQ Standby would have
to be deleted and installed again."
> 
> Newton Alex wrote:
>     I think ""This alert will trigger if HAWQ Standby is not synchronized with HAWQ Master.
To synchronize HAWQ Standby with HAWQ Master, use the service action 'Re-Synchronize HAWQ
Standby' on HAWQ service page." is good enough.
>     The other sentence "If HAWQ Standby is too far behind HAWQ Master, HAWQ Standby would
have to be deleted and installed again." can be part of the Resync action itself. It need
not be part of this alert.

Newton, that is a great idea.
Considering that Re-Sync is a custom action, where do you think we should show this message?
On the stdout of Ambari console? Would have been great if it we had a pop-up after the action,
like a reminder.


- Matt


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/43753/#review119802
-----------------------------------------------------------


On Feb. 18, 2016, 7:33 p.m., Matt wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/43753/
> -----------------------------------------------------------
> 
> (Updated Feb. 18, 2016, 7:33 p.m.)
> 
> 
> Review request for Ambari, Alexander Denissov, bhuvnesh chaudhary, Goutam Tadi, jun aoki,
Lav Jain, and Newton Alex.
> 
> 
> Bugs: AMBARI-15104
>     https://issues.apache.org/jira/browse/AMBARI-15104
> 
> 
> Repository: ambari
> 
> 
> Description
> -------
> 
> Update the description for the alert so that the user knows what to do if the alert is
triggered from the description on the alert page
> 
> 
> Diffs
> -----
> 
>   ambari-server/src/main/resources/common-services/HAWQ/2.0.0/alerts.json 3119a0c 
> 
> Diff: https://reviews.apache.org/r/43753/diff/
> 
> 
> Testing
> -------
> 
> Manually Tested. Descriptions appears on the alerts page.
> 
> No additional unit tests required.
> 
> 
> Thanks,
> 
> Matt
> 
>


Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message