FlowSphera 2.0.0 build 3 release notes

After a power cut the database could stay offline — ClickHouse sets aside the parts it had not finished writing to the disk when the power went, but refused to open a table with more than 100 of them, and the collector waited for it until someone repaired it by hand (seen: 18 hours without records). Such parts are now set aside however many there are, so the database always comes up by itself. FT_CH_FSYNC=1 in /etc/flowsphera/env also has the records written to the disk before they count (fewer lost at a power cut, inserts about 10× slower; off by default).

Event «Database recovered after an unclean stop» — on the Events page and on every notification channel (Telegram, e-mail, webhook, SIEM): when the server stopped, which tables lost parts, whether the newest records are missing, the time without records, and from when the totals for long periods were summed again from the records (they are, automatically, so long and short periods agree). Parts set aside are removed after 30 days.