Mobile loop detected on a newly configured Outlook in LAN

maxxer

Hi.
Few days ago we installed a new PC and configured Outlook with ActiveSync against Z-Push with Zimbra backend. The first sync took rather long time, so I went to the log files and there are a lot of Mobile loop detected.

Shouldn’t this be caused by connection errors? How can they be, if the client is in LAN?

Thanks

fbartels

Hi @maxxer,

you can read more about loop detection and causes for loops at https://wiki.z-hub.io/display/ZP/Loop+detection.

maxxer

So I should enable debug to understand why Outlook re-requested so many items?

fbartels

As you can read on the linked page:

these are just examples. Z-Push would ignore an appointment with “a start date after the end date” even before sending it to the mobile, because it’s semantically broken. For the user it’s the same, the item is not on the mobile. The reason will be stated in the output of z-push-admin.

I think the reason for an item to not be synced is also logged into the z-push error logging.

liverpoolfcfan

Outlook tries to download 512 items at a time. If your z-push and or zimbra are on servers that are any way slow, and particularly if you have a lot of logging turned up, the z-push data collection from zimbra can take so long that the client gives up and immediately issues the same request again. This can leave the original z-push thread still collecting data while a second one starts in parallel. This cycle can repeat and choke up your servers. This issue should no longer be a problem for the most part with some of the throttling logic added last year.

You could further alleviate this possibility by reducing SYNC_MAX_ITEMS - try 256 or 128 to see if it makes any difference. Note - at steady state this max has almost no impact. It is only during an initial sync of a mailbox where it increases the number or round trips needed.

Another possibility is that there are so many items in one of the folders that the initial enumeration of them for the sync states is taking too long.

If you do turn up logging, make sure to set the ZIMBRA_DEBUG to the specific user so you don’t tax the server with logging for everyone

maxxer

@liverpoolfcfan thank you very much for the detailed explaination.

This server is the busiest we have, there are ~90 users and 130 AS sessions (not all active). Most of them are with normal handeld devices, but there is a percentage of Outlook clients, which are the most problematic ones.

Anyway the server load is very low, always lower than 0.3 and rarely going higher. Using ZWC is fast, so I don’t understand why it’s sometimes slow on AS replies. The infrastructure is not really up to date: Zimbra 8.6, PHP 5.5, ZP 2.3.9, backend 67.

I’m going to migrate state files to MySQL to see if it helps in improving performance.