Github user cshannon commented on the issue:
https://github.com/apache/activemq-artemis/pull/1853
Had to update the PR again as something got messed up when rebasing
---
Github user cshannon commented on the issue:
https://github.com/apache/activemq-artemis/pull/1853
@clebertsuconic - I did some cleanup work and I added a couple of basic
tests to the compatibility test. It doesn't do much but show that the current
snapshot can load a 2.4 journal succ
Github user cshannon commented on the issue:
https://github.com/apache/activemq-artemis/pull/1853
@clebertsuconic - The metric in the page counters is persisted along with
the page counters now. When the page counters are written to the journal that
metric is encoded to disk and relo
Github user clebertsuconic commented on the issue:
https://github.com/apache/activemq-artemis/pull/1853
@cshannon although looking at the PR now you added persistent size into
page counters... but as I read it, it is not persistent...
We should go all the way in.. i.e...
Github user cshannon commented on the issue:
https://github.com/apache/activemq-artemis/pull/1853
Oh ok I get it now. Yeah you are right that the PagingStore interface has
a method to get the number of pages and the page size which can be used to get
an idea of consumption. This val
Github user clebertsuconic commented on the issue:
https://github.com/apache/activemq-artemis/pull/1853
I was just saying paging metrics should be exposed. Even if the files are
to be GCed but if something happened and they are not the user should have the
metrics exposed about pages.
Github user cshannon commented on the issue:
https://github.com/apache/activemq-artemis/pull/1853
@clebertsuconic - I don't think we need to re-calculate anything if the
values are missing for paging. The point of this metric isn't to give the
user information on storage consumption
Github user clebertsuconic commented on the issue:
https://github.com/apache/activemq-artemis/pull/1853
I think paging should be measured just in number of pages (files) on the
address. That would give users an idea about storage consumption. It
wouldnât indeed be practical to load
Github user cshannon commented on the issue:
https://github.com/apache/activemq-artemis/pull/1853
I updated the PR with getPersistentSize() everywhere along with comments to
try and make it more clear what the value represented. I can tweak it some
more if people still think it's not
Github user cshannon commented on the issue:
https://github.com/apache/activemq-artemis/pull/1853
Ok I can just use getPersistentSize() then for everything, I will update
the PR shortly.
---
Github user clebertsuconic commented on the issue:
https://github.com/apache/activemq-artemis/pull/1853
I would include the sum you mentioned for large message. The large message
file size plus the large message encode size. That translates fine as
getPersjstentsjze
---
Github user cshannon commented on the issue:
https://github.com/apache/activemq-artemis/pull/1853
@clebertsuconic - hmm so I'm starting to like you original idea of just
using getEncodeSize() more as the the message might be non-durable and in that
case the encode size is still used
Github user cshannon commented on the issue:
https://github.com/apache/activemq-artemis/pull/1853
getPersistentSize() is fine with me as it specifies that the size
represents the entire persisted size of the message
---
Github user clebertsuconic commented on the issue:
https://github.com/apache/activemq-artemis/pull/1853
@cshannon you may not like getEncodeSize().. but doing
message.getMessageSize() doesn't make it any clearer..we need to find a better
name? maybe getPersistentSize()?
Github user cshannon commented on the issue:
https://github.com/apache/activemq-artemis/pull/1853
@clebertsuconic - Fair enough, but I think I should call it
getMessageSize() instead (i already did this in a couple places) The reason
why I would use that name is because the value is
Github user clebertsuconic commented on the issue:
https://github.com/apache/activemq-artemis/pull/1853
i would rename getSize() as getEncodeSize() on the interfaces then. getSize
could be dubious and users may get the wrong expectation of the meaning.
---
Github user cshannon commented on the issue:
https://github.com/apache/activemq-artemis/pull/1853
@clebertsuconic - No, I'm not trying to measure the size on the JVM. I
want to measure the size of the messages in general (including how much space
is on disk) as a metric to know how m
Github user clebertsuconic commented on the issue:
https://github.com/apache/activemq-artemis/pull/1853
You are getting the getEncodeSize() as the size... what is what you want to
measure? how much memory it uses on the JVM?
We have some calculations we do on paging.. perhaps
Github user cshannon commented on the issue:
https://github.com/apache/activemq-artemis/pull/1853
@michaelandrepearce - ok made those 2 changes so it is ready for another
look
---
Github user cshannon commented on the issue:
https://github.com/apache/activemq-artemis/pull/1853
@michaelandrepearce - I updated my PR, take a look and let me know what you
think.
---
Github user cshannon commented on the issue:
https://github.com/apache/activemq-artemis/pull/1853
This PR is quite large so I would appreciate if multiple people take a look
to make sure it looks ok.
---
21 matches
Mail list logo
