summaryrefslogtreecommitdiffstats
path: root/docs/development/design/inspector-design-guideline.rst
diff options
context:
space:
mode:
authorYujun Zhang <zhang.yujunz@zte.com.cn>2017-09-05 20:11:42 +0800
committerYujun Zhang <zhang.yujunz@zte.com.cn>2017-09-05 20:13:25 +0800
commitf5a750f19896f6f161876a5babc5b42c720dedf8 (patch)
tree0afa505cd59721d16f9ca69d98e432efded2adf6 /docs/development/design/inspector-design-guideline.rst
parent1c1d3d1ddc21ae28deabef9f5f6e310d46fdf0ef (diff)
Add host specific VM list to inspector design guide line
JIRA: DOCTOR-73 Change-Id: I163b97ec87761538594034e926055646223af327 Signed-off-by: Yujun Zhang <zhang.yujunz@zte.com.cn>
Diffstat (limited to 'docs/development/design/inspector-design-guideline.rst')
-rw-r--r--docs/development/design/inspector-design-guideline.rst19
1 files changed, 18 insertions, 1 deletions
diff --git a/docs/development/design/inspector-design-guideline.rst b/docs/development/design/inspector-design-guideline.rst
index 4add8c0f..faa5e424 100644
--- a/docs/development/design/inspector-design-guideline.rst
+++ b/docs/development/design/inspector-design-guideline.rst
@@ -31,7 +31,24 @@ Guidelines
Host specific VMs list
----------------------
-TBD, see `DOCTOR-76`_.
+While requirement in doctor project is to have alarm about fault to consumer in one second, it is just a limit we have
+set in requirements. When talking about fault management in Telco, the implementation needs to be by all means optimal
+and the one second is far from traditional Telco requirements.
+
+One thing to be optimized in inspector is to eliminate the need to read list of host specific VMs from Nova API, when
+it gets a host specific failure event. Optimal way of implementation would be to initialize this list when Inspector
+start by reading from Nova API and after this list would be kept up-to-date by ``instance.update`` notifications
+received from nova. Polling Nova API can be used as a complementary channel to make snapshot of hosts and VMs list in
+order to keep the data consistent with reality.
+
+This is enhancement and not perhaps something needed to keep under one second in a small system. Anyhow this would be
+something needed in case of production use.
+
+This guideline can be summarized as following:
+
+- cache the host VMs mapping instead of reading it on request
+- subscribe and handle update notifications to keep the list up to date
+- make snapshot periodically to ensure data consistency
Parallel execution
------------------