diff options
Diffstat (limited to 'book-enea-nfv-core-release-info/doc/known_bugs_and_limitations.xml')
-rw-r--r-- | book-enea-nfv-core-release-info/doc/known_bugs_and_limitations.xml | 233 |
1 files changed, 71 insertions, 162 deletions
diff --git a/book-enea-nfv-core-release-info/doc/known_bugs_and_limitations.xml b/book-enea-nfv-core-release-info/doc/known_bugs_and_limitations.xml index bc682ab..1aa394d 100644 --- a/book-enea-nfv-core-release-info/doc/known_bugs_and_limitations.xml +++ b/book-enea-nfv-core-release-info/doc/known_bugs_and_limitations.xml | |||
@@ -44,29 +44,35 @@ | |||
44 | </listitem> | 44 | </listitem> |
45 | 45 | ||
46 | <listitem> | 46 | <listitem> |
47 | <para>Instances fail to boot when using a direct port (SR-IOV) on | 47 | <para>ThunderX integrated NICs cannot be used for SR-IOV</para> |
48 | ThunderX</para> | ||
49 | 48 | ||
50 | <itemizedlist> | 49 | <itemizedlist> |
51 | <listitem> | 50 | <listitem> |
52 | <para>Description and Impact:</para> | 51 | <para>Description and Impact:</para> |
53 | 52 | ||
54 | <para>Deployment is successful with SR-IOV enabled interfaces | 53 | <para>For the moment ENEA NFV Core is missing the support to |
55 | however, instances fail to boot when a direct bound (SR-IOV) port | 54 | configure ThunderX integrated NICs for deployment. Furthermore, |
56 | is added. This has been tested using a SR-IOV capable PCI Express | 55 | ThunderX integrated NICs cannot be used for SR-IOV even if |
57 | Network Interface. As a consequence it is impossible to | 56 | configured manually after deployment. This happens because |
58 | passthrough a SR-IOV port on ThunderX.</para> | 57 | ThunderX integrated NICs are themseleves virtual functions and are |
58 | incorrectly handled by libvirt when trying to assign them to a | ||
59 | virtual machine.</para> | ||
60 | |||
61 | <para>It is however possible to deploy with SR-IOV over add-on | ||
62 | interfaces via the PCI-E expansion slots.</para> | ||
59 | </listitem> | 63 | </listitem> |
60 | 64 | ||
61 | <listitem> | 65 | <listitem> |
62 | <para>Workaround: N/A.</para> | 66 | <para>Workaround: there is no workaround for this issue. As an |
67 | alternative, the user can configure an external PCI-E NIC for | ||
68 | SR-IOV.</para> | ||
63 | </listitem> | 69 | </listitem> |
64 | </itemizedlist> | 70 | </itemizedlist> |
65 | </listitem> | 71 | </listitem> |
66 | 72 | ||
67 | <listitem> | 73 | <listitem> |
68 | <para>Security groups are not working correctly for ICMP traffic in | 74 | <para>ThunderX integrated NICs cannot be used for PCI passthrough with |
69 | deployments with OpenDaylight.</para> | 75 | direct-physical bound Neutron ports</para> |
70 | 76 | ||
71 | <itemizedlist> | 77 | <itemizedlist> |
72 | <listitem> | 78 | <listitem> |
@@ -74,52 +80,32 @@ | |||
74 | 80 | ||
75 | <itemizedlist> | 81 | <itemizedlist> |
76 | <listitem> | 82 | <listitem> |
77 | <para>When OPNFV is deployed with OpenDaylight as an SDN | 83 | <para>PCI Passthrough using direct-physical bound ports also |
78 | controller, the Security Groups rules pertaining to ICMP do | 84 | uses the neutron-sriov-agent. Because the interfaces are |
79 | not work as expected. The OpenFlow rules describing the ICMP | 85 | represented as virtual functions, it will be impossible to use |
80 | rules are inconsistent, so VMs can be pinged even when this is | 86 | Neutron ports bound as direct-physical (the Nova driver will |
81 | not desired.</para> | 87 | identify them as type-VF, not type-PF).</para> |
82 | </listitem> | 88 | |
83 | 89 | <para>Due to this, it is not possible to claim PCI devices | |
84 | <listitem> | 90 | using direct-physical bound ports.</para> |
85 | <para>This reproduces on aarch64. On x86 the security groups | ||
86 | work correctly.</para> | ||
87 | </listitem> | 91 | </listitem> |
88 | </itemizedlist> | 92 | </itemizedlist> |
89 | </listitem> | ||
90 | |||
91 | <listitem> | ||
92 | <para>Workaround: N/A.</para> | ||
93 | </listitem> | ||
94 | </itemizedlist> | ||
95 | </listitem> | ||
96 | |||
97 | <listitem> | ||
98 | <para>Virtual instances do not get IP from DHCP in SFC scenarios with | ||
99 | ODL</para> | ||
100 | |||
101 | <itemizedlist> | ||
102 | <listitem> | ||
103 | <para>Description and Impact:</para> | ||
104 | 93 | ||
105 | <itemizedlist> | 94 | <itemizedlist> |
106 | <listitem> | 95 | <listitem> |
107 | <para>After a fresh deploy of OPNFV with OpenDaylight and the | 96 | <para>It is however possible to passthrough any device using |
108 | Service Function Chaining scenario configurations, instances | 97 | the PCI alias method, which requries configuring a whitelist |
109 | fail to get IP from DHCP, due to OpenDaylight | 98 | of PCI devices and assigning an alias which is set as metadata |
110 | malfunctioning.</para> | 99 | in the Nova flavor.</para> |
111 | </listitem> | ||
112 | |||
113 | <listitem> | ||
114 | <para>The SFC VNFs are not reachable via SSH for management | ||
115 | and configuration.</para> | ||
116 | </listitem> | 100 | </listitem> |
117 | </itemizedlist> | 101 | </itemizedlist> |
118 | </listitem> | 102 | </listitem> |
119 | 103 | ||
120 | <listitem> | 104 | <listitem> |
121 | <para>Workaround: Restarting OpenDaylight via | 105 | <para>Workaround:</para> |
122 | <literal>systemctl</literal> fixes the problem.</para> | 106 | |
107 | <para>There is no workaround for this issue. As an alternative, | ||
108 | the user can configure a PCI alias instead.</para> | ||
123 | </listitem> | 109 | </listitem> |
124 | </itemizedlist> | 110 | </itemizedlist> |
125 | </listitem> | 111 | </listitem> |
@@ -242,106 +228,6 @@ | |||
242 | </listitem> | 228 | </listitem> |
243 | 229 | ||
244 | <listitem> | 230 | <listitem> |
245 | <para>Fuel Healthcheck Stack update test fails</para> | ||
246 | |||
247 | <itemizedlist> | ||
248 | <listitem> | ||
249 | <para>Description and Impact:</para> | ||
250 | |||
251 | <para>The Platform test case number 5 (Update stack) from the Fuel | ||
252 | Healthcheck sometimes fails. This has no impact on the overal | ||
253 | cluster functionality.</para> | ||
254 | </listitem> | ||
255 | |||
256 | <listitem> | ||
257 | <para>Workaround: N/A.</para> | ||
258 | </listitem> | ||
259 | </itemizedlist> | ||
260 | </listitem> | ||
261 | |||
262 | <listitem> | ||
263 | <para>Issue #1 with Openstack Resource Agents and Compute Fencing | ||
264 | functionality</para> | ||
265 | |||
266 | <itemizedlist> | ||
267 | <listitem> | ||
268 | <para>Description and Impact:</para> | ||
269 | |||
270 | <para>In an OPNFV deployment that uses Openstack Resource Agents, | ||
271 | the neutron-openvswitch-agent is killed by Pacemaker when booting, | ||
272 | due to Pacemaker misconfiguration.</para> | ||
273 | </listitem> | ||
274 | |||
275 | <listitem> | ||
276 | <para>Workaround:</para> | ||
277 | |||
278 | <para>Starting the <literal>systemd</literal> service manually | ||
279 | makes it run successfully. Enea NFV Core 1.0.1 is shipped without | ||
280 | Openstack Resource Agents, therefore this issue should not affect | ||
281 | the user.</para> | ||
282 | </listitem> | ||
283 | </itemizedlist> | ||
284 | </listitem> | ||
285 | |||
286 | <listitem> | ||
287 | <para>Issue #2 with Openstack Resource Agents and Compute Fencing | ||
288 | functionality</para> | ||
289 | |||
290 | <itemizedlist> | ||
291 | <listitem> | ||
292 | <para>Description and Impact:</para> | ||
293 | |||
294 | <para>In an OPNFV deployment that uses Openstack Resource Agents, | ||
295 | when we configure the <literal>fence_compute</literal> as a | ||
296 | Pacemaker resource, the Controller nodes start to reboot each | ||
297 | other endlessly.</para> | ||
298 | </listitem> | ||
299 | |||
300 | <listitem> | ||
301 | <para>Workaround:</para> | ||
302 | |||
303 | <para>Enea NFV Core 1.0.1 is shipped without Openstack Resource | ||
304 | Agents, therefore this issue should not affect the user.</para> | ||
305 | </listitem> | ||
306 | </itemizedlist> | ||
307 | </listitem> | ||
308 | |||
309 | <listitem> | ||
310 | <para>Virtual instances are not affected by removing a node from the | ||
311 | Ceph Storage Cluster</para> | ||
312 | |||
313 | <itemizedlist> | ||
314 | <listitem> | ||
315 | <para>Description and Impact:</para> | ||
316 | |||
317 | <para>Engineering wanted to validate the survival of storage | ||
318 | systems when a single disk is removed, without causing data loss. | ||
319 | Without physical access to the test setup, this test is not | ||
320 | feasible.</para> | ||
321 | </listitem> | ||
322 | |||
323 | <listitem> | ||
324 | <para>Workaround:</para> | ||
325 | |||
326 | <itemizedlist> | ||
327 | <listitem> | ||
328 | <para>The chosen approach was to validate what happens to the | ||
329 | Ceph cluster, when network connectivity is lost for the | ||
330 | Storage interface of one of the nodes. No impact was observed | ||
331 | when running an instance using Ceph for volume storage.</para> | ||
332 | </listitem> | ||
333 | |||
334 | <listitem> | ||
335 | <para><ulink | ||
336 | url="http://ceph.com/geen-categorie/admin-guide-replacing-a-failed-disk-in-a-ceph-cluster/">Reference | ||
337 | information</ulink></para> | ||
338 | </listitem> | ||
339 | </itemizedlist> | ||
340 | </listitem> | ||
341 | </itemizedlist> | ||
342 | </listitem> | ||
343 | |||
344 | <listitem> | ||
345 | <para>Offline Deploy with Fuel fails at times</para> | 231 | <para>Offline Deploy with Fuel fails at times</para> |
346 | 232 | ||
347 | <itemizedlist> | 233 | <itemizedlist> |
@@ -394,44 +280,67 @@ | |||
394 | </listitem> | 280 | </listitem> |
395 | 281 | ||
396 | <listitem> | 282 | <listitem> |
397 | <para>Fuel Healthcheck Stack creation with wait condition test, | 283 | <para>On Mixed Arch Deployment, only the aarch64 TestVM Cirros image |
398 | fails</para> | 284 | will be installed by Fuel</para> |
399 | 285 | ||
400 | <itemizedlist> | 286 | <itemizedlist> |
401 | <listitem> | 287 | <listitem> |
402 | <para>Description and Impact:</para> | 288 | <para>Description and Impact:</para> |
403 | 289 | ||
404 | <para>The platform test case (create stack with wait condition) | 290 | <para>Due to the fact that Fuel will only deploy the aarch64 |
405 | from the Fuel Healthcheck, fails. This has no impact on overall | 291 | image, Yardstick, Functest, and certain Health Check tests will |
406 | cluster functionality.</para> | 292 | not work. These test suites are dependent on a single image name |
293 | at a time, and do not know on how to place instances on the | ||
294 | Compute for images that each require a different arch.</para> | ||
295 | |||
296 | <para>To have both testVM images, the user must add the x86_64 | ||
297 | image manually.</para> | ||
407 | </listitem> | 298 | </listitem> |
408 | 299 | ||
409 | <listitem> | 300 | <listitem> |
410 | <para>Workaround: N/A.</para> | 301 | <para>There is no workaround for the test suites failures.</para> |
411 | </listitem> | 302 | </listitem> |
412 | </itemizedlist> | 303 | </itemizedlist> |
413 | </listitem> | 304 | </listitem> |
414 | 305 | ||
415 | <listitem> | 306 | <listitem> |
416 | <para>On Mixed Arch Deployment, only the aarch64 TestVM Cirros image | 307 | <para>Removing QoS policies is unreliable</para> |
417 | will be installed by Fuel</para> | ||
418 | 308 | ||
419 | <itemizedlist> | 309 | <itemizedlist> |
420 | <listitem> | 310 | <listitem> |
421 | <para>Description and Impact:</para> | 311 | <para>Description and Impact:</para> |
422 | 312 | ||
423 | <para>Due to the fact that Fuel will only deploy the aarch64 | 313 | <para>When removing per port bandwidth limiting QoS policies, all |
424 | image, Yardstick, Functest, and certain Health Check tests will | 314 | traffic is suddenly dropped. On the the other hand, when removing |
425 | not work. These test suites are dependent on a single image name | 315 | QoS policies configured at Openstack network level, traffic flows |
426 | at a time, and do not know on how to place instances on the | 316 | as if the rules are still there.</para> |
427 | Compute for images that each require a different arch.</para> | 317 | </listitem> |
428 | 318 | ||
429 | <para>To have both testVM images, the user must add the x86_64 | 319 | <listitem> |
430 | image manually.</para> | 320 | <para>There is no workaround.</para> |
321 | </listitem> | ||
322 | </itemizedlist> | ||
323 | </listitem> | ||
324 | |||
325 | <listitem> | ||
326 | <para>Enabling Ceph for Glance and Nova ephemeral storage makes the | ||
327 | deployment fail on aarch64</para> | ||
328 | |||
329 | <itemizedlist> | ||
330 | <listitem> | ||
331 | <para>Description and Impact:</para> | ||
332 | |||
333 | <para>There are multiple configurable Storage Backends in Fuel | ||
334 | settings. Enabling Ceph RBD for images (Glance) and Ceph RBD for | ||
335 | ephemeral volumes (Nova), makes the deployment fail at the CEPH | ||
336 | Ready Check performed on the primary Controller node. This only | ||
337 | occurs when using aarch64 nodes; on x86_64, deployment does not | ||
338 | fail</para> | ||
431 | </listitem> | 339 | </listitem> |
432 | 340 | ||
433 | <listitem> | 341 | <listitem> |
434 | <para>There is no workaround for the test suites failures.</para> | 342 | <para>There is no workaround. The user should not enable these |
343 | options.</para> | ||
435 | </listitem> | 344 | </listitem> |
436 | </itemizedlist> | 345 | </itemizedlist> |
437 | </listitem> | 346 | </listitem> |