v1.4.0

New Features

• EXPERIMENTAL: Add support of Apple M1 devices (#1122, @terasum @AstroProfundis @sunxiaoguang)
  • Playground may not fully work as some components don't yet have packages for darwin-arm64 released
• Not displaying dashboard address if it's "none" or "auto" (#1054, @9547)
• Support filtering nodes and roles in check subcommand of tiup-cluster (#1030, @AstroProfundis)
• Support retry of failed operations from where it broke with replay command of tiup-cluster and tiup-dm (#1069 #1157, @lucklove)
• Support upgrade and patch a stopped TiDB / DM cluster (#1096, @lucklove)
• Support setting global custom values for topology of tiup-cluster (#1098, @lucklove)
• Support custom root_url and anonymous login for Grafana in tiup-cluster (#1085, @mianhk)
• Support remote read and remote write for Prometheus node in tiup-cluster (#1070, @XSHui)
• Support custom external AlertManager target for Prometheus node in tiup-cluster (#1149, @lucklove)
• Support force reinstallation of already installed component (#1145, @9547)
• Add --force and retain data options to tiup-dm (#1080, @9547)
• Add enable/disable subcommands to tiup-dm (#1114, @9547)
• Add template subcommand to tiup-cluster to print pre-defined topology templates (#1156, @lucklove)
• Add --version option to display subcommand of tiup-cluster to print the cluster version (#1207, @AstroProfundis)
• Allow value type change when editing topology with edit-config subcommand of tiup-cluster (#1050, @AstroProfundis)

Fixes

• Not allowing deployment if the input topology file is empty (#994, @AstroProfundis)
• Fix data dir setting for Prometheus (#1040, @9547)
• Fix the issue that pre-defined Prometheus rules may be missing if a custom rule_dir is set (#1073, @9547)
• Fix the issue that config files of Prometheus and Grafana are not checked before start (#1074, @9547)
• Fix the issue that cluster name is not validated for some operations (#1177, @AstroProfundis)
• Fix the issue that tiup-cluster reloads a cluster even if the config may contain errors (#1183, @9547)
• Fix the issue that publish command may fail when uploading files without retry (#1174 #1202, @AstroProfundis; #1167, @lucklove)
• Fix the issue that newly added TiFlash nodes may fail to start during scale-out in tiup-cluster (#1227, @9547)
• Fix incorrect cluster name in alert messages (#1238, @9547)
• Fix the issue that blackbox_exporter may not collecting ping metrics correctly (#1250, @STRRL)

Improvements

• Reduce jitter during upgrade process of TiDB cluster
  • Make sure PD node is online and serving before upgrading the next one (#1032, @HunDunDM)
  • Upgrade PD leader node after upgrading other PD nodes (#1086, @AstroProfundis)
  • Increase schedule limit during upgrade of TiKV nodes (#1661, @AstroProfundis)
  • Add check to validate if all regions are healthy (#1126, @AstroProfundis)
• Only reload Prometheus configs when needed (#989, @9547)
• Show default option on prompted input messages (#1132 #1134, @wangbinhe3db)
• Include user's input in error message if prompted challenge didn't pass (#1104, @AstroProfundis)
• Check for data_dir and log_dir overlap before deploying a cluster (#1093, @9547)
• Improve checking rules in tiup cluster check command (#1099 #1107, @AstroProfundis; #1118 #1124, @9547)
• Refine list and display command for tiup-cluster (#1139, @baurine)
• Mark patched nodes in display output of tiup-cluster and tiup-dm (#1125, @AstroProfundis)
• Ignore users.* settings for TiFlash if the cluster version is later than v4.0.12 and v5.0.0-rc (#1211, @JaySon-Huang)
• Cache timestamp manifest in memory to reduce network requests (#1212, @lucklove)
• Upgrade toolchain to Go 1.16 (#1151 #1153 #1130, @AstroProfundis)
• Use GitHub Actions to build and release TiUP components (#1158, @AstroProfundis)
• Remove deprecated v0manifest support, TiUP version before v1.0.0 may not be able to download latest packages anymore (#906)