lp://staging/~simeon-blanc/+junk/sync_opti_offset_nofake_bf

Related source package recipes

No recipes using this branch. (?)

Create packaging recipe

Related snap packages

Related bugs

Related blueprints

629. By sbc_msf on 2016-01-25

SP-224 [IMP] Turn on the puller cache and sort updates according to their id

This is important now because we rely on the last ID to fetch the updates. If they
 are not sorted correctly, we could skip some updates.

This change has not impact on their execution order because we already sort them
 before execution.

628. By sbc_msf on 2016-01-21

SP-224 [IMP] Add comments according to Fabien's comments

627. By sbc_msf on 2016-01-21

SP-224 [IMP] Fix an issue involving the offset used when recovering from a crash

The offset wasn't correctly set when receiving a packet from the SYNC_SERVER. As
 a result, after a crash, we fetched the same updates several times. This is not
 a major issue because an update should be idempotent.

626. By sbc_msf on 2016-01-20

SP-224 [IMP] Change the way the offset is computed during a sync

The offset was already computed correctly but any new filter on server-side
 might have corrupted this offset.

Now, the server gives the offset. As a result, it can define how the instance
 should increment it.

625. By sbc_msf on 2016-01-18

SP-224 [IMP] Remove a check that is useless thanks to Fabien

This check was at the wrong place because a crash would have occurred before
 if the condition wasn't true.

This commit moves the check so that it is done before any crash.

624. By sbc_msf on 2016-01-08

SP-224 [IMP] Set the correct sequence number when sending updates from sync to instances

This bug was added in the source code while developping a new server-side algorithm
 to group the updates before sending them to the instances.

The "master update" used as an example to group similare updates together has a
 sequence number. This sequence number was not taken into account when grouping it
 with other updates.

As a result, the instances fetched the updates with a wrong sequence number.

623. By sbc_msf on 2016-01-07

SP-224 [IMP] Remove the fake packet

The pulling phase is quick enough so that we don't need any fake
 update. Removing it make the code clearer and prevent our instances
 from executing fake updates.

622. By sbc_msf on 2016-01-06

SP-224 [IMP] Add a filter on the rows to keep the offset as low as possible

This commit allow an instance to fetch the updates based on a sequence number
(the last sequence number already fetched by the instance) and an offset (as
before) or an update id.

This functionnality relies on the fact the updates are sorted according to their
 to their ID first, and then their sequence number. According to the code, this
 fact might be wrong in production.

621. By sbc_msf on 2016-01-05

SP-224 [IMP] Add 500 updates in each packet

Before this commit, Unifield didn't put 500 updates in each packet event
if the packet size was set to 500.

620. By jftempo on 2015-12-21

Rules v1.124: set master data on updates