Problem Description -
Ceilometer does not provide Instance-level network tx/rx errors and drops.
It appears that the Ceilometer compute agent does not poll some Instance network meters. These metrics are available via the nova diagnostics command for each Instance.
GET /v2/{tenant_id}/servers/{server_id}/diagnostics
provides the following -
<diagnostics>
<tap138c86fc-55_tx_errors>0</tap138c86fc-55_tx_errors>
<vda_write>258048</vda_write>
<cpu0_time>13020000000</cpu0_time>
<tap138c86fc-55_rx_packets>21</tap138c86fc-55_rx_packets>
<tap138c86fc-55_tx>2660</tap138c86fc-55_tx>
<tap138c86fc-55_rx_errors>0</tap138c86fc-55_rx_errors>
<tap138c86fc-55_rx>2502</tap138c86fc-55_rx>
<vda_read>15964672</vda_read>
<vda_write_req>66</vda_write_req>
<memory-actual>524288</memory-actual>
<tap138c86fc-55_rx_drop>0</tap138c86fc-55_rx_drop>
<tap138c86fc-55_tx_packets>23</tap138c86fc-55_tx_packets>
<memory>524288</memory>
<memory-rss>117052</memory-rss>
<vda_read_req>918</vda_read_req>
<vda_errors>-1</vda_errors>
<tap138c86fc-55_tx_drop>0</tap138c86fc-55_tx_drop>
</diagnostics>
This request is to make the following network and disk error meters available via Ceilometer -
tx_errors
rx_errors
tx_drop
rx_drop
Environment
OpenStack Mitaka release
Raising a new bug for vda_errors (vbd_errors, etc. for each disk present). Updating description of this bug to include tx/rx errors and drop.