Bug #3525
closedhandover is unreliable
0%
Description
when attempting intra-cell handovers using sysmoBTS, I currently get at best a success rate of 1 out of 10.
This tracks investigation of the problem.
Updated by neels over 5 years ago
- Category deleted (
osmo-bts-sysmo) - Status changed from New to In Progress
- Priority changed from Normal to High
Measurement reports weirdness:
- Normally, when activating a channel when establishing a call, we very soon get measurements for the new channel and send a measurement report to the BSC.
- After RSL Chan Activ and RSL Chan Activ ACK for handover, however, there are no measurements for a while.
Using a not-yet-merged patch from pmaier to log measurements in detail ("adding measurement") in osmo-bts, I see that we are not receiving any measurements from the l1.
Looking at the failing cases:
0ms Channel Activ Ack (no measurements from l1 all this time) 700ms Handover Failure
Looking at the cases where handover succeeds:
0ms Channel Activ Ack 200ms first Handover Detection couple more Handover Detection 250ms last Handover Detection 260ms first measurement from l1, from now on every ~15ms 400ms RSL Establish IND 550ms Handover Complete
Looks bleak in comparison to a normal TCH/F channel activation:
0ms Chan Activ ACK 2ms first measurement from l1 21ms 2nd 45ms 3rd 56ms 4th 62ms 5th
(these timings according to the packet timestamps when the gsmtap_log is received on the core host, didn't bother with the gsmtap_log timestamps themselves)
It actually looks like the l1 purposefully waits for a Handover Detection before it even considers sending measurements.
Which does make sense to me, as these are uplink measurements, and we wouldn't get any if the MS isn't telling us about them, IIUC.
So, I think examining measurements for handover, as I did now, is actually moot.
Updated by neels over 5 years ago
- Status changed from In Progress to Rejected
so I assumed my BTS were well calibrated, but after going through sysmobts-calib the handover is super reliable. 100% success rate.
(sigh) another episode of "how to waste time unnecessarily by assuming things".
Updated by laforge over 5 years ago
On Wed, Sep 05, 2018 at 01:22:21PM +0000, neels [REDMINE] wrote:
- Normally, when activating a channel when establishing a call, we very soon get measurements for the new channel and send a measurement report to the BSC.
- After RSL Chan Activ and RSL Chan Activ ACK for handover, however, there are no measurements for a while.
Using a not-yet-merged patch from pmaier to log measurements in detail ("adding measurement") in osmo-bts, I see that we are not receiving any measurements from the l1.
This may be related to the L1 SAPIs we activate at that point. I guess we only activate RACH, then wait for the RACH bursrts to arrive, and then actually activate the TCH/F, FACCH/F, SACCH/TF (for a TCH/H) SAPIS. And then my suspicion is that measurement reports are only generated once those L1 SAPIs are active.
Probably not worth investigating further as you have discovered BTS frequency calibration was to be blamed.