action #10584
Updated by okurz about 8 years ago
## observation tests are still failing sometimes in bootloader_s390 due to unexpected behaviour when downloading kernel and initrd... ## steps to reproduce * "clone_job.pl" on any s390x@zVM job often enough, at least 20 times, to see the issue ## problem H1. REJECTED: too many jobs started at once cause some timeout on download --> random back-off time in qaboot did not remedy all problems, reverted changes H2. ACCEPTED: Network download can stall some times and we can't control this --> ~~undecided undecided if this is true, still ~~ "retry" as workaround applied, see #10584#note-18 H3. REJECTED: Network download is just slower sometimes, causing progress notification to be output more often, causing one buffer window of x3270 to fill up *and* the function `expect_s3270` in the backend is not handling this correctly. --> expect_s3270 *was* behaving correctly just having wrong comments and misleading output, see #10584#note-21 H4. ## suggestion * ~~some tasks have been tried, workarounds applied, what is left to do:~~ * DONE: Fix handling of "status line matches neither buffer_ready nor buffer_full:" in `expect_s3270` to accept either H3 or H2 and reject the other -> expect_s3270 was actually fine but the message "status line matches neither buffer_ready nor buffer_full" was very confusing and improved in [gh-osautoinst#514](https://github.com/os-autoinst/os-autoinst/pull/514) ## further details the former ftpboot fix (#10224) (poo#10224) seems to be not the only issue with that so i will investigate this further