Attention is currently required from: neels.
11 comments:
Patchset:
Hi Neels,
Thanks for the extensive review. I have addressed many of your individual comments but not yet made the switch to your recommended approach for increasing accuracy. I will submit another update tomorrow. For now, I wanted to clean up the noise on this changeset so far and clarify its purpose.
Thanks again,
-Michael
File include/osmocom/bsc/gsm_data.h:
Patch Set #8, Line 1130: unsigned long long gsm_lchan_active_duration_ms(const struct gsm_lchan *lchan);
(idea: use uint64_t for a fixed size independent from arch. […]
Done
File src/osmo-bsc/bts.c:
Patch Set #8, Line 1514: "Cummulative number of active milliseconds on TCH chans",
("Cumulative") […]
I agree the name is messy. However, this will indeed hold the cumulative number of milliseconds, not the average duration per activation. I am deriving the average based on this total in the VTY. I am also planning to use this total to show Erlangs per site.
Perhaps "Cumulative number of milliseconds of TCH channel activity."?
File src/osmo-bsc/bts_trx_vty.c:
Patch Set #8, Line 574: uint duration_s = (gsm_lchan_active_duration_ms(lchan) + 500) / 1000;
the +500 acts as a round(), yet the reported descriptions below suggest that floor() should be used […]
Done
File src/osmo-bsc/bts_vty.c:
Patch Set #8, Line 3826: uint64_t activations_tch = rate_ctr_group_get_ctr(bts->bts_ctrs, BTS_CTR_CHAN_ACT_TCH)->current;
osmocom style dictates that local variables are declared at the start of the function (or scope)
Done
File src/osmo-bsc/gsm_data.c:
Patch Set #8, Line 348: /* Get duration of active time for this lchan in milliseconds */
Rather: […]
Done
Patch Set #8, Line 352: if (lchan->activate.concluded) {
osmocom style asks for early-exit: […]
Done
Patch Set #8, Line 357: duration = elapsed.tv_sec * 1000LL + elapsed.tv_nsec / 1000000;
the LL on the 1000 has no effect, the first operand defines the integer size for calculation. […]
Done
File src/osmo-bsc/lchan_fsm.c:
You should rather put this bit in a new lchan_st_established_onleave() function. […]
This is a great bit of information. I do think it is important to capture the error timeouts. If we see SDCCH averages trending toward a long timeout value, we know something is up. I've moved the stat increment code and log to lchan_fsm_unused_onenter.
Patch Set #8, Line 1811: /* Add active milliseconds to cummulative counts per channel type */
/* Report how long this lchan was active, to add to the average activity duration metric. […]
Adopted. However, I will use "cumulative duration" to reflect that stat's purpose as mentioned before.
Patch Set #8, Line 1817: LOG_LCHAN(lchan, LOGL_INFO, "GSM_LCHAN_TCH was active for %llu milliseconds\n", duration_ms);
i would rather just have a single LOG_LCHAN() that uses gsm_chan_t_name() above the switch, because […]
Done
To view, visit change 27081. To unsubscribe, or for help writing mail filters, visit settings.