Releases: CS-SI/SafeScale
SafeScale v20.06.1
New
Tenants file
- Added the ability to set metadata Bucket Name to store SafeScale metadata (dynamically generated by default) using keyword
Bucket
inmetadata
section of tenants.{toml|json|yml} file
Fixed bugs
- Gateway failover doesn't work; FIXED
- Helm was installed on cluster expand even if disabled at cluster creation; FIXED
- Virtual IP doesn't work with RedHat/CentOS 8; FIXED
- Minimum Image size not honored at host creation; FIXED
- Feature docker does not work as expected with RedHat/CentOS 7; FIXED
- Fails to create a cluster of flavour K8S with complexity normal or large and hardening disabled; FIXED
- Feature remotedesktop does not work as expected in RedHat/CentOS; FIXED
- Various outscale driver issues; FIXED
- Various cleanup issues when removing Virtual IP; FIXED
- Fail to correctly cleanup cluster when created with flag -k and when the creation failed; FIXED
- Panic if no tenants file is found; FIXED
- gRPC panics; FIXED
safescale ssh tunnel
does not create a fully functional SSH tunnel; FIXED
SafeScale v20.06.0
New
SafeScale Infra:
- added Cloud Provider OutScale support
- added support of RedHat/CentOS 8
SafeScale Platform
safescale cluster kubectl
now accept paramter-f -
(reading content from standard input)
Changes
SafeScale Infra
- added a way to define internal DNS suffix of host using
--domain
parameter ofsafescale host create
command - deprecated support of Debian 8 (may work in the future, but will not be tested anymore)
- deprecated support of Ubuntu 17.10 and 18.10 (may work in the future, but will not be tested anymore)
SafeScale Platform
- added a way to define internal DNS suffix for all hosts created in a cluster, using
--domain
parameter ofsafescale cluster create
command
Build
- go v1.14 works correctly (1.15 is not tested)
- minimum go version to compile SafeScale is now v1.12
Fixed bugs
- the result of
safescale cluster inspect
shows remotedesktop URL(s) even if remotedesktop has been disabled: fixed safescale
/safescaled
panics if$HOME/.safescale
folder does not exist: fixed- not possible to ssh on gateway from master or node, with account
cladm
: fixed - in feature,
sfHelm repo add
fails with error--tls unknown parameter
: fixed - when cluster is deleted, mounted volumes lead to operation failure with corrupted metadata: now an error message is displayed saying the volumes have to be detached first
- with huaweicloud stack (used by FlexibleEngine and OpenTelekom providers), sizing of system disk size is not honoured: fixed
Known bugs
- VirtualIP does not work with RedHat/CentOS 8
SafeScale v20.03.2
Bugfixes:
- Issue #81: Problem when start/stop an host, command fails
- timeout issue in package lib/client leading to out of order safescale bucket commands
SafeScale v20.03.1
Bugfix:
- Issue #80: If cluster expand fails, subsequent calls of cluster expand fail on "host not found" error
SafeScale v20.03.0
New
SafeScale Infra
- Added AWS support
- Added LAN Virtual IP support for OVH provider. Note that the required functionality on OVH side, Floating IP, is in beta at the time writing and must be asked explicitely to your OVH account manager
- Removed the CloudWatt driver, the provider having closed shop
Binaries
- First release of macOS 64bits binaries
- First release of Raspberry Pi 32bits binaries (armv5)
Changes
OVH provider:
- OVH deprecated KeyStone v2.0, v3 is now used
- Object Storage region namings have changed: no more index in region name (ie. SBG3 becomes SBG).
safescale tenant set
will output an error describing the change that has to be made in the tenant configuration file.
Important notice about next release
The next version will no longer contain support for SWARM, DCOS and OHPC cluster flavors. From this version on, efforts will focus only on BOH (Bunch Of Hosts) and K8S (Kubernetes).
In the indefinite future, it is possible that these flavors will be reintroduced as plugins for SafeScale; however, it is unlikely that these flavors will be maintained by the CS Group team, so contributors will have to be found for this.
If a particular flavor is needed for your project, express yourself through the github number to trace the need.
SafeScale v20.01.0
New:
SafeScale Infra:
safescale ssh connect
: added ability to set username and/or shell to usesafescale ssh run
: displays outputs as soon as they come (previously everything was displayed at the end of the command)
SafeScale Platform:
K8S cluster flavor:
- version 1.14.1 used by default
- kubernetes is now hardened by default. If you want to disable this hardening at cluster creation, use the parameter
--disable hardening
- helm v2, installed by default, can be disabled, using
--disable helm
- In Normal or Large mode, K8S cluster is configured in HA mode
Globally:
- The reverse proxy, installed by default in a cluster, can be disabled, using
--disable reverseproxy
- The remote desktop, installed by default on master(s), can be disabled, using
--disable remotedesktop
- The failover of gateways installed by default when complexity is
Normal
orLarge
can be disabled, using--disable gateway-failover
Changes:
safescale data
removed. Maybe will come back later as it's own cli.
Bugfixes:
- Explicitly set SSH parameter PasswordAuthentication to no (don't rely on distributions)
safescale
command panics if $HOME/.safescale doesn't exist- Overall stability improved with better error handling (far less risk of panic)
Contributing:
- minimum go version to build SafeScale is now 1.12
- prepared move to go modules
- added
docs/CODING_GUIDELINES.md
to centralize the coding guidelines adopted for SafeScale - added
make style
to test compliance to coding guidelines adopted
SafeScale v19.10.0
New:
-
Gateway HA:
When the provider supports it (currently pure Openstack, FlexibleEngine and OpenTelekom), it's now possible to create a network with 2 gateway hosts and a Virtual IP which behaves as default route for the network. One gateway hosts the VIP at a time, and is endorsed by the other gateway host of the first one fails.
In case of a cluster, in complexities normal and large, this ability is used automatically (it can be disabled).
When the provider used doesn't support Virtual IP, safescaled will provude a warning log indicating the lack of support but will continue will a single gateway host. -
new feature : postgresxl
Installs on a cluster a Postgres-XL service on master(s). Postgres-XL is a multi-master database environment, that provides almost all the features of PostgreSQL (one notable exception being subtransactions...). -
New feature : geoserver
Installs a GeoServer instance on the cluster (for now without HA)
Changes:
- Better error handling (the goal is to get rid of panics everywhere, still work in progress)
- Proxy rules in feature: supports upstreams parameters, to change algorithm and related settings of Kong
Fixes:
- Detected and fixed data races
Documentation:
We are definitely late on the documentation, we will publish a revision soon with an update of it.
SafeScale v19.06.1
Fixed bug #72: scanner now loads properly provider drivers to scan the tenants marked Scannable = true
SafeScale v19.06.0
News:
- !!BETA!! driver
gcp
for Google Cloud Platform. Considered beta in this release, some parts of the driver may change in the near future (in particular the parameters to use in tenants definition file, currently being as provided by GCP, but not following SafeScale formalism) K8S
(kubernetes) cluster flavor is restored (after being suspended in releases v19.03.x) and is based on the last stable version of Kubernetes at cluster creationDCOS
cluster flavor is restored (after being suspended in releases v19.03.x) and is based on DCOS v1.13- In tenant definition file :
- added keyword
WhitelistTemplateRegexp
to define a list of white-listed host templates using regexp - added keyword
BlacklistTemplateRegexp
to define a list of black-listed host templates using regexp - added keyword
WhitelistImageRegexp
to define a list of white-listed host images using regexp - added keyword
BlacklistImageRegexp
to define a list of black-listed host images using regexp - added keyword
AvailabilityZone
to set the availability zone wanted for the tenant. This keyword is mandatory for almost all the providers (notable exception being OVH which doesn't have multiple availability zones at the time being)
- added keyword
- In
safescale host create
, added parameter--sizing
in replacement of--cpu
,--ram
,--disk
(which are still present but deprecated); documentation has not been updated on that point, please seesafescale host create --help
for details - In
safescale cluster create
, added--sizing
,--gw-sizing
,--master-sizing
,--node-sizing
in replacement of--cpu
,--ram
,--disk
(which are still present but deprecated); documentation has not been updated on that point, please seesafescale host create --help
for details
Changes:
- Kong (used as reverse proxy for clusters) upgraded to v1.1
- Command
safescale
results are now consistently returning json data, except forsafescale ssh connect
andsafescale ssh run
- Operator account used by SafeScale on remote hosts has been changed from
gpac
tosafescale
. To continue to use resources created with v19.03, you can use the keywordOperatorUsername: gpac
incompute
section of tenants description file - feature description file:
- added the ability to target gateway when a feature is installed on a cluster
- added the ability to define a default value for a parameter
Fixes:
- Various data races found and fixed in
safescale
andsafescaled
- Coherent availability zone is now enforced through tenants file; this ensures all resources are created in the same availability zone (not guaranteed in previous versions)
- Network creation ensures CIDR are RFC1918-compliant; it's not possible to ask for routable CIDR anymore
Documentation:
- Added doc about feature (usage and file description)
- Overall documentation update (but not complete yet)