Bug 14078
| Summary: | pinger function in nanny.c in the piranha package | ||
|---|---|---|---|
| Product: | [Retired] Red Hat High Availability Server | Reporter: | Jeremy Rumpf <jrumpf> |
| Component: | piranha | Assignee: | Phil Copeland <copeland> |
| Status: | CLOSED ERRATA | QA Contact: | Phil Copeland <copeland> |
| Severity: | medium | Docs Contact: | |
| Priority: | medium | ||
| Version: | 1.0 | CC: | jrumpf |
| Target Milestone: | --- | ||
| Target Release: | --- | ||
| Hardware: | i386 | ||
| OS: | Linux | ||
| Whiteboard: | |||
| Fixed In Version: | Doc Type: | Bug Fix | |
| Doc Text: | Story Points: | --- | |
| Clone Of: | Environment: | ||
| Last Closed: | 2000-07-18 15:43:14 UTC | Type: | --- |
| Regression: | --- | Mount Type: | --- |
| Documentation: | --- | CRM: | |
| Verified Versions: | Category: | --- | |
| oVirt Team: | --- | RHEL 7.3 requirements from Atomic Host: | |
| Cloudforms Team: | --- | Target Upstream Version: | |
| Embargoed: | |||
This problem has been found and new RPMs will be posted shortly Thanks! This has been fixed and will be posted shortly. When the updated RPMs are released, you will be able to download them from http://people.redhat.com/kbarrett/HA/ , and they will be announced on piranha-list. New RPMs have been posted. 0.4.16-3 |
I tried to get this on the piranha mailing list but couldn't get a response from the server. I figured this would be the best alternative. I'm posting this using the piranha source from the piranha-0.4.16-2.src.rpm. If you setup pulse in the lvs.cf file with a send statement and no expect statement, example virtual egovnetwww1 { address = XXX.XXX.XXX.XXX eth1:1 #external addy of #the vserver active = 1 load_monitor = rup #Use rstatd to #determine the load timeout = 20 #Seconds before #a server is considered #dead reentry = 40 #Seconds we wait when #a server comes back #online port = 80 #External/internal #listening port protocol = tcp #self explanitory scheduler = wlc #Weighted least connection send = "\n" #expect = "" #These are the test strings #to be sent and recieved #to determine if the #service is alive #These will default for http persistent = 180 #Maintain a persistent #connections for 180 seconds pmask = 255.255.255.255 #Actual physical server #definitions follow etc, etc.... The pinger function will fail waiting for the expect string even though it is not required. This causes the removal of the server from the virtual server. A simple fix is in order (though I am in now way an expert). The original around line 340 of the nanny.c file is as follows: 340 if (i) { 341 if (expect_str) { 342 read_len = strlen(expect_str); 343 if (read_len >= sizeof(read_buf)) 344 read_len = sizeof(read_buf) - 1; 345 346 i = (int) read(connectSock, read_buf, read_len); 347 348 if (i >= 0) { 349 read_buf[i] = 0; 350 351 if (flags & NANNY_FLAG_VERBOSE) { 352 piranha_log(flags, 353 (char *) "read expected len=%d, text=\"%s\"", 354 strlen(expect_str), expect_str); 355 piranha_log(flags, (char *) "read got len=%d, text=%s", 356 strlen(read_buf), read_buf); 357 } 358 359 if ((read_len == 1) && (*expect_str == '*')) { 360 read_buf[0] = '*'; 361 read_buf[1] = 0; 362 i = 1; 363 } /* Wildcard */ 364 365 if (i >= read_len) { 366 if (strncmp((const char *) expect_str, 367 (const char *) read_buf, read_len)) 368 return_status = 1; 369 } 370 else 371 return_status = 1; 372 } 373 else 374 return_status = 1; 375 } 376 } 377 378 close(connectSock); 379 380 if (!i) { 381 piranha_log(flags, (char *) "read from %s:%d timed out", 382 inet_ntoa(*whereto), port); 383 return_status = 1; 384 } 385 386 return return_status; 387 } Changed to: 340 if (i) { 341 if (expect_str) { 342 read_len = strlen(expect_str); 343 if (read_len >= sizeof(read_buf)) 344 read_len = sizeof(read_buf) - 1; 345 346 i = (int) read(connectSock, read_buf, read_len); 347 348 if (i >= 0) { 349 read_buf[i] = 0; 350 351 if (flags & NANNY_FLAG_VERBOSE) { 352 piranha_log(flags, 353 (char *) "read expected len=%d, text=\"%s\"", 354 strlen(expect_str), expect_str); 355 piranha_log(flags, (char *) "read got len=%d, text=%s", 356 strlen(read_buf), read_buf); 357 } 358 359 if ((read_len == 1) && (*expect_str == '*')) { 360 read_buf[0] = '*'; 361 read_buf[1] = 0; 362 i = 1; 363 } /* Wildcard */ 364 365 if (i >= read_len) { 366 if (strncmp((const char *) expect_str, 367 (const char *) read_buf, read_len)) 368 return_status = 1; 369 } 370 else 371 return_status = 1; 372 } 373 else 374 return_status = 1; 375 } 376 377 378 if (!i) { 379 piranha_log(flags, (char *) "read from %s:%d timed out", 380 inet_ntoa(*whereto), port); 381 return_status = 1; 382 } 383 } 384 385 close(connectSock); 386 return return_status; 387 } Just put the test for the read time out within exect_string stanza and place the close command before the return. It seems, it still waits for the poll timeout but ignores everything since expect_str is NULL. I came across this trying to cluster some nonstandard servers (ie banner servers) where the default HTTP request didn't apply. When trying to override the default http send and expect strings manually in the lvs.cf file nanny still reported read timeouts when the send was specified without the expect. Hope this helps..... jrumpf