stratos-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Sajith Kariyawasam <saj...@wso2.com>
Subject Re: [Discuss] Suggesting a patching model for PCA
Date Tue, 01 Dec 2015 07:20:26 GMT
Hi Chamila,

If the whole PCA is patched there is no way to differentiate whether its a
patch or a new released version, which would not easy to maintain I guess.
In a patch fix, IMO those kind of file removal / refactoring could be very
minimal. We can expect those changes in a version upgrade.
But even if we come across such special cases where some files are removed
/ renamed in a patch fix, I would like to propose to include a script with
the patch itself, which would do the file replace / removal etc.



On Mon, Nov 30, 2015 at 11:41 PM, Chamila De Alwis <chamilad@wso2.com>
wrote:

> Hi Sajith,
>
> If we are going to handle each file separately wouldn't it introduce
> complications when files are moved and refactored? Wouldn't it be better to
> patch the PCA as a whole rather than file by file?
>
>
> Regards,
> Chamila de Alwis
> Committer and PMC Member - Apache Stratos
> Software Engineer | WSO2 | +94772207163
> Blog: code.chamiladealwis.com
>
>
>
> On Mon, Nov 30, 2015 at 6:44 PM, Sajith Kariyawasam <sajith@wso2.com>
> wrote:
>
>>
>> On Sat, Nov 28, 2015 at 11:55 PM, Imesh Gunaratne <imesh@apache.org>
>> wrote:
>>
>>>
>>>
>>> On Tue, Nov 24, 2015 at 12:58 PM, Sajith Kariyawasam <sajith@wso2.com>
>>> wrote:
>>>>
>>>>
>>>> There are two aspects in here as I see.
>>>>  1) Applying patches to already running server and restart.
>>>>  2) Applying patches to a 'fresh' server and start
>>>>
>>>> It would have been ideal if we can achieve both scenarios above, but
>>>> yes, with the points you have mentioned (1) is not possible. I remember we
>>>> had a discussion in past regarding this,
>>>> did we finalized it?. So, if there is a running cluster (say,100 nodes)
>>>> and if we need to apply a patch, approach would be to spawn up 'new'
>>>> instances with the patches?
>>>>
>>>> Yes, currently we might need to do this because, if we apply a patch in
>>> an active instance and restart the server. During that period it will
>>> receive requests via the load balancer and those will fail. Few months back
>>> Shaheed proposed a software update management feature to handle this logic.
>>>
>>>
>>>> Even in that scenario, there is no easy way to track the changes being
>>>> done to the originally released binary, since the current approach is to
>>>> replace the old files with patched files.
>>>> The model I was thinking is to implement a model where PCA always keep
>>>> the original files (backed up) and read the latest files (from a patch
>>>> location), which is the exact way of Carbon
>>>> patch model.
>>>>
>>>> Right, I think this is a good approach. Can you please explain it using
>>> an example?
>>>
>>>
>> I have attached a sample directory structure. There will be a separate
>> directory named "patches", into which the patches need to be copied.
>> Say we have patched "agentgithandler.py" in patch0500 and its copied in
>> patches directory as shown.
>> When the agent starts up it checks whether there are any "patchxxxx"
>> directories, and if so it backs up the original file(s) into patch0000, and
>> replace original file with the patched file.
>>
>> If the patch folder is removed we need to restore the original file, in
>> order to cater for that, the logic should be, at the startup agent checks
>> whether there are any files in patch0000 directory
>> (which means a patch is already applied) and search for that file in
>> patchxxxx directories, if not found it compares the md5sum with the file in
>> the original location.
>>  If it doesn't match, it will replace with the file in patch0000.
>>
>> Those logic needs to be written in python code and to be executed before
>> starting any of the agent related code.
>>
>> I will try to send a PR soon so that it will be much clear about the
>> process, and we can review it and do any modifications needed.
>>
>>
>>
>>
>>> Thanks
>>>
>>>
>>>> Can you please explain the patching process you are proposing?
>>>>>
>>>>> Thanks
>>>>>
>>>>> On Wed, Nov 11, 2015 at 6:07 PM, Sajith Kariyawasam <sajith@wso2.com>
>>>>> wrote:
>>>>>
>>>>>> Hi Devs,
>>>>>>
>>>>>> At the moment Python Cartridge Agent (PCA) doesn't have a convenient
>>>>>> way of applying patches.
>>>>>> If there is an issue identified and fixed, in order to apply the
fix
>>>>>> to a running system that particular file(s) has to be replaced in
the
>>>>>> running system,
>>>>>> and there is no easy way to revert the patch is there is any issue
>>>>>> found later.
>>>>>> Also, there is no way of tracking the list of patches applied.
>>>>>>
>>>>>> I would like to suggest to implement a patching model for PCA,
>>>>>> similar to that we have in Stratos core, which is based on carbon
[1]
>>>>>> WDYT ?
>>>>>>
>>>>>> [1]
>>>>>> https://docs.wso2.com/display/Carbon420/WSO2+Patch+Application+Process
>>>>>>
>>>>>> Thanks,
>>>>>> Sajith
>>>>>>
>>>>>> --
>>>>>> Sajith Kariyawasam
>>>>>> *Committer and PMC member, Apache Stratos, *
>>>>>> *WSO2 Inc.; http://wso2.com <http://wso2.com>*
>>>>>> *Mobile: 0772269575 <0772269575>*
>>>>>>
>>>>>
>>>>>
>>>>>
>>>>> --
>>>>> Imesh Gunaratne
>>>>>
>>>>> Senior Technical Lead, WSO2
>>>>> Committer & PMC Member, Apache Stratos
>>>>>
>>>>
>>>>
>>>>
>>>> --
>>>> Sajith Kariyawasam
>>>> *Committer and PMC member, Apache Stratos, *
>>>> *WSO2 Inc.; http://wso2.com <http://wso2.com>*
>>>> *Mobile: 0772269575 <0772269575>*
>>>>
>>>
>>>
>>>
>>> --
>>> Imesh Gunaratne
>>>
>>> Senior Technical Lead, WSO2
>>> Committer & PMC Member, Apache Stratos
>>>
>>
>>
>>
>> --
>> Sajith Kariyawasam
>> *Committer and PMC member, Apache Stratos, *
>> *WSO2 Inc.; http://wso2.com <http://wso2.com>*
>> *Mobile: 0772269575*
>>
>
>


-- 
Sajith Kariyawasam
*Committer and PMC member, Apache Stratos, *
*WSO2 Inc.; http://wso2.com <http://wso2.com>*
*Mobile: 0772269575*

Mime
View raw message