A watched game is left without a standby for minutes after each swap #64

Closed
opened 2026-10-31 10:32:31 -04:00 by cmoriarty · 3 comments
Owner

What happens

The card often reads "Standby: Looking for one", most noticeably right after a swap, which is when the operator is looking. A swap or a recovery takes the standby (resolution.take_standby), and the slot stays empty until the scheduler verifies another link. That takes minutes, because:

  • Reads are paced. One site read at a time across the whole app (READ_CONCURRENCY 1), in rounds 60 s apart (PREWARM_INTERVAL_SECONDS), each trying at most five links (WALK_LINKS_PER_READ). On 2026-10-31 reads averaged 35 s.
  • Most links are dead. In the event log from 2026-10-26 to 2026-10-31, 137 of 595 links tried verified (23%).
  • The same dead links lead every read. On Chelsea vs. Man Utd, givemestreams.cfd, buff-streams.buzz and resports.cfd answered "no stream found" on every check from noon to full time. They sort first by tier and host score, so after each 10-minute cooldown they took three of the five tries again. The sunk rule (reputation.is_sunk) doesn't catch them: it judges a host across all games, and these hosts play other games (givemestreams.cfd and resports.cfd played Lakers vs. Warriors on 2026-10-30).
  • Failed links cool for 10 minutes. A game with four or five working hosts, one of them playing, runs out once two of the others have failed.

Seen live, 2026-10-31, Chelsea vs. Man Utd (UTC)

Time What happened Standby
13:12 tuned on iohit.st shd247.world ready
13:19 the operator's swap took it empty 30 s, until topsurge.st verified
13:20 a degrading swap took that empty 4 min (one five-link read found nothing)
13:40 a degrading swap, then a recovery, took the next two empty 1 min, then 2 min
13:43–14:22 iohit.st was re-read six times; shd247.world stayed ready ready
14:22 a recovery took it at full time empty to the end: iohit.st and shd247.world cooling, topsurge.st refused for speed (#63), the rest "no stream found"

A standby was ready about 90% of the time from the tune to 14:22. Every gap followed a swap or a recovery that took it.

The ideas

  1. A second standby for a played game. Keep two verified standbys for any game being played, so a swap still leaves one ready. The extra reads are only for played games, which already go ahead of readiness (PRIORITY_STANDBY). The standby re-check (every 150 s, against the CDN, not a site read) covers both. The followed Ready games keep one standby each, as decided on 2026-10-29.
  2. Demote a link that keeps finding nothing for this game. After a few "no stream found" answers for the same game (three, say), sort that link after the game's other fresh links, as a sunk host is: still tried, but last. This is per game and per link, so a host that works for other games keeps its place there. Each five-link read then spends its tries on links that might work.

Possibly also: start the refill as soon as a swap takes the standby, rather than at the next round. It's a smaller gain; the 30 s and 1 min gaps were mostly the read itself.

Not part of this: why those hosts find no stream for this game. Either the hubs list hosts that don't carry it, or a player our reader doesn't recognise.

## What happens The card often reads "Standby: Looking for one", most noticeably right after a swap, which is when the operator is looking. A swap or a recovery takes the standby (`resolution.take_standby`), and the slot stays empty until the scheduler verifies another link. That takes minutes, because: - **Reads are paced.** One site read at a time across the whole app (`READ_CONCURRENCY` 1), in rounds 60 s apart (`PREWARM_INTERVAL_SECONDS`), each trying at most five links (`WALK_LINKS_PER_READ`). On 2026-10-31 reads averaged 35 s. - **Most links are dead.** In the event log from 2026-10-26 to 2026-10-31, 137 of 595 links tried verified (23%). - **The same dead links lead every read.** On Chelsea vs. Man Utd, givemestreams.cfd, buff-streams.buzz and resports.cfd answered "no stream found" on every check from noon to full time. They sort first by tier and host score, so after each 10-minute cooldown they took three of the five tries again. The sunk rule (`reputation.is_sunk`) doesn't catch them: it judges a host across all games, and these hosts play other games (givemestreams.cfd and resports.cfd played Lakers vs. Warriors on 2026-10-30). - **Failed links cool for 10 minutes.** A game with four or five working hosts, one of them playing, runs out once two of the others have failed. ## Seen live, 2026-10-31, Chelsea vs. Man Utd (UTC) | Time | What happened | Standby | |---|---|---| | 13:12 | tuned on iohit.st | shd247.world ready | | 13:19 | the operator's swap took it | empty 30 s, until topsurge.st verified | | 13:20 | a degrading swap took that | empty 4 min (one five-link read found nothing) | | 13:40 | a degrading swap, then a recovery, took the next two | empty 1 min, then 2 min | | 13:43–14:22 | iohit.st was re-read six times; shd247.world stayed ready | ready | | 14:22 | a recovery took it at full time | empty to the end: iohit.st and shd247.world cooling, topsurge.st refused for speed (#63), the rest "no stream found" | A standby was ready about 90% of the time from the tune to 14:22. Every gap followed a swap or a recovery that took it. ## The ideas 1. **A second standby for a played game.** Keep two verified standbys for any game being played, so a swap still leaves one ready. The extra reads are only for played games, which already go ahead of readiness (`PRIORITY_STANDBY`). The standby re-check (every 150 s, against the CDN, not a site read) covers both. The followed Ready games keep one standby each, as decided on 2026-10-29. 2. **Demote a link that keeps finding nothing for this game.** After a few "no stream found" answers for the same game (three, say), sort that link after the game's other fresh links, as a sunk host is: still tried, but last. This is per game and per link, so a host that works for other games keeps its place there. Each five-link read then spends its tries on links that might work. **Possibly also:** start the refill as soon as a swap takes the standby, rather than at the next round. It's a smaller gain; the 30 s and 1 min gaps were mostly the read itself. **Not part of this:** why those hosts find no stream for this game. Either the hubs list hosts that don't carry it, or a player our reader doesn't recognise.
Author
Owner

Decided 2026-10-31: both ideas are in the batch.

  • The second standby is for played games only. Followed Ready games keep one standby each, as decided on 2026-10-29.
  • Demotion is per game and per link: after three "no stream found" answers for the same game, the link sorts after that game's other fresh links. It is still tried, like a sunk host.
  • The early refill, starting the next standby read as soon as a swap takes one, goes in if it stays small.
**Decided 2026-10-31:** both ideas are in the batch. - **The second standby is for played games only.** Followed Ready games keep one standby each, as decided on 2026-10-29. - **Demotion is per game and per link:** after three "no stream found" answers for the same game, the link sorts after that game's other fresh links. It is still tried, like a sunk host. - **The early refill,** starting the next standby read as soon as a swap takes one, goes in if it stays small.
Author
Owner

Deployed 2026-10-31 17:44 UTC (PR #67, 1938e97). Left open until a second standby is seen holding.

Seen live on a test tune of Spurs vs. C Palace (17:45–17:52 UTC):

  • The refill after a swap. The channel stalled and swapped to iohit.st at 17:47:45. A read for a new standby ran straight after it, tried five links and found nothing (17:48:45). The next read verified powerstream.club at 17:49:03: 78 s from swap to a new standby.
  • The second-standby reads ran but found nothing. The read at 17:51:39 tried five more links. Its "no stream found" hosts (givemestreams, buff-streams, s2watch, resports) are the disguised player of #68, so on games like this a second standby mostly waits on #68.

Done when a played channel's card shows "then ".

**Deployed 2026-10-31 17:44 UTC (PR #67, 1938e97). Left open until a second standby is seen holding.** **Seen live on a test tune of Spurs vs. C Palace (17:45–17:52 UTC):** - **The refill after a swap.** The channel stalled and swapped to iohit.st at 17:47:45. A read for a new standby ran straight after it, tried five links and found nothing (17:48:45). The next read verified powerstream.club at 17:49:03: 78 s from swap to a new standby. - **The second-standby reads ran but found nothing.** The read at 17:51:39 tried five more links. Its "no stream found" hosts (givemestreams, buff-streams, s2watch, resports) are the disguised player of #68, so on games like this a second standby mostly waits on #68. **Done when** a played channel's card shows "then <host>".
Author
Owner

Closed on the tests at the operator's word (2026-11-01): reopen, or file a new ticket, if production shows a problem.

Closed on the tests at the operator's word (2026-11-01): reopen, or file a new ticket, if production shows a problem.
Sign in to join this conversation.
No labels
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference
cmoriarty/sports-breaker#64
No description provided.