| Summary | Activesync Ping Request Limit |
| Queue | Synchronization |
| Queue Version | Git master |
| Type | Enhancement |
| State | Rejected |
| Priority | 1. Low |
| Owners | mrubinsk (at) horde (dot) org |
| Requester | horde (at) albasoft (dot) com |
| Created | 08/20/2013 (4340 days ago) |
| Due | |
| Updated | 08/20/2013 (4340 days ago) |
| Assigned | |
| Resolved | 08/20/2013 (4340 days ago) |
| Milestone | |
| Patch | No |
State ⇒ Rejected
5.0.4". I'll update and try again to confirm behaviour.
I can see the changes in current code you pointed in your comment.
Thank you
Assigned to Michael Rubinsky
State ⇒ Feedback
Horde_ActiveSync_Collections::pollForChanges(). We store the starting
timestamp of each PING or Heartbeat SYNC request in the syncCache,
then on each iteration of the PING/SYNC loop we verify that the
timestamp of the currently running sync is not older than the
timestamp stored in the syncCache (which would indicate another PING
process has started). With this code, there should never be more than
2 running PING requests at anytime, and the older request should
detect that it's stale and die.
If this is not working for you, I would need to see the synclog
covering the requests that this is failing for.
FWIW I wanted to use a semaphore to monitor this, but the shared
memory functions in PHP needed to do this are not present on every
system. Horde_Cache could be used, but there is no guarantee that one
is available. At some point, I'll probably end up sniffing out the
available capabilities then fall back to the syncCache method though
this will add complexity.
Priority ⇒ 1. Low
Type ⇒ Enhancement
Summary ⇒ Activesync Ping Request Limit
Queue ⇒ Synchronization
Milestone ⇒
Patch ⇒ No
State ⇒ New
server. In addition, those requests have a very long lasting live,
compared to any other web server request.
Current Activesync code has no limit to the number of Ping requests a
server is going to attend for a single user/device. One of my devices
(Android phone) had some issue that made it fire a Ping request every
few seconds, so the server ended up trying to fulfill more than 90
concurrent Ping requests, which degraded whole server performance. It
shouldn't be a problem if those requests weren't so long lasting.
I think there's no point in attending more than 2 or 3 concurrent Ping
requests for a single user/device, beyond dealing with mobile devices
when switching between wireless connections. Right now an error on a
single device can freeze a nice production server in just a few minutes.