summaryrefslogtreecommitdiffstats
path: root/rubbos/app/tomcat-connectors-1.2.32-src/native/TODO.txt
diff options
context:
space:
mode:
authorhongbotian <hongbo.tianhongbo@huawei.com>2015-11-30 02:41:33 -0500
committerhongbotian <hongbo.tianhongbo@huawei.com>2015-11-30 02:43:36 -0500
commit9401f816dd0d9d550fe98a8507224bde51c4b847 (patch)
tree94f2d7a7893a787bafdca8b5ef063ea316938874 /rubbos/app/tomcat-connectors-1.2.32-src/native/TODO.txt
parente8ec7aa8e38a93f5b034ac74cebce5de23710317 (diff)
upload tomcat
JIRA: BOTTLENECK-7 Change-Id: I875d474869efd76ca203c30b60ebc0c3ee606d0e Signed-off-by: hongbotian <hongbo.tianhongbo@huawei.com>
Diffstat (limited to 'rubbos/app/tomcat-connectors-1.2.32-src/native/TODO.txt')
-rw-r--r--rubbos/app/tomcat-connectors-1.2.32-src/native/TODO.txt371
1 files changed, 371 insertions, 0 deletions
diff --git a/rubbos/app/tomcat-connectors-1.2.32-src/native/TODO.txt b/rubbos/app/tomcat-connectors-1.2.32-src/native/TODO.txt
new file mode 100644
index 00000000..73dd13f3
--- /dev/null
+++ b/rubbos/app/tomcat-connectors-1.2.32-src/native/TODO.txt
@@ -0,0 +1,371 @@
+ Licensed to the Apache Software Foundation (ASF) under one or more
+ contributor license agreements. See the NOTICE file distributed with
+ this work for additional information regarding copyright ownership.
+ The ASF licenses this file to You under the Apache License, Version 2.0
+ (the "License"); you may not use this file except in compliance with
+ the License. You may obtain a copy of the License at
+
+ http://www.apache.org/licenses/LICENSE-2.0
+
+ Unless required by applicable law or agreed to in writing, software
+ distributed under the License is distributed on an "AS IS" BASIS,
+ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ See the License for the specific language governing permissions and
+ limitations under the License.
+
+TODO for tomcat-connectors
+
+$Id: TODO.txt 920112 2010-03-07 20:57:35Z timw $
+
+1) Optimize "distance"
+======================
+
+Sorting the list of balanced workers by distance would be nice, but:
+How to combine the sorting with the offset implementation (especially
+useful for strategy BUSYNESS under low load).
+
+Local error states and likely other features will also break in case
+we do naive reordering.
+
+2) Reduce number of string comparisons in most_suitable
+========================================================
+
+a) redirect/domains
+
+It would be easy to improve the redirect string b an integer, giving the
+index of the worker in the lb. Then lb would not need to search for the redirect worker.
+
+The same way, one could add a list with indizes to workers in the same domain.
+Whenever domain names are managed (init and status worker update) one would
+scan the worker list and update the index list.
+
+Finally one could have a list of workers, whose domain is the same as the redirect
+attribute of the worker, because that's also something we consider.
+
+What I'm not sure about, even in the existing code, is the locking between updates
+by the status worker and the process local information about the workers,
+especially in the case, when status updates a redirect or domain attribute.
+
+I would like to keep these attributes and the new index arrays process local,
+and the processes should find out about changes made by status to shm (redirect/domain)
+and then rebuild their data. No need to get these on every request from the shm,
+only the check for up-to-date should be made.
+
+b) exact matches for jvmRoutes
+
+Could we use hashes instead of string comparisons all the time?
+I'm not sure, if a good enough hash takes longer than a string comparison though.
+
+3) Code separation between factory, validate and init in lb
+============================================================
+
+The factory contains:
+
+ private_data->worker.retries = JK_RETRIES;
+ private_data->s->recover_wait_time = WAIT_BEFORE_RECOVER;
+
+I think, this should move to validate() or init().
+It might even be obsolete, because in init, we already have:
+
+ pThis->retries = jk_get_worker_retries(props, p->s->name,
+ p->s->retries = pThis->retries;
+ p->s->recover_wait_time = jk_get_worker_recover_timeout(props, p->s->name, WAIT_BEFORE_RECOVER);
+ if (p->s->recover_wait_time < WAIT_BEFORE_RECOVER)
+ p->s->recover_wait_time = WAIT_BEFORE_RECOVER;
+
+Then: In validate there is
+
+ p->lb_workers[i].s->error_time = 0;
+
+So shouldn't there also be
+
+ p->lb_workers[i].s->maintain_time = time(NULL);
+
+4) Logging
+==========
+
+a) Allow logging of request url or uuid in jk log to ease matching with access log.
+
+b) Implement log rotation for IIS. (done in 1.2.31)
+
+c) Allow adding of log notes for IIS like we do with Apache.
+
+d) Add error type info to access log notes
+
+e) Refactor: Use the same code files for the request logging functions in Apache 1.3 and 2.0.
+
+f) Refactor: Use the same code files for piped logging in Apache 1.3 and 2.0.
+
+5) ajpget
+==========
+
+Combine ajplib and Apache ab to an ajp13 commandline client ajpget.
+
+6) Parsing workers.properties
+=============================
+
+Parsing of workers.properties aditionally to just looking up attributes
+would help users to detect syntax errors in the file. At the moment
+no information will be logged, e.g. when attributes contain typos.
+
+Example: worker.list vs. worker.lists.
+
+7) Persisting workers.properties
+================================
+
+Make workers.properties persist from inside status worker.
+
+Add additional properties file, that contains a journal of property changes done
+via the status worker. When initializing those overwrite the initial workers.properties.
+
+Update actions in the status worker will allow to optionally add a change to this journal.
+We can also add a comment with timestamp etc. to each journal line.
+
+8) Reduce number of uses of time(NULL)
+======================================
+
+We use time(NULL) a lot. Since it only has resolution of a second,
+I'm asking myself, if we could update the actual time in only a few
+places and get time out of some variables when needed. The same does
+not hold true for millisecond time, but in several cases we use the time,
+it's not very critical, that it is exact. These cases are related to:
+
+Some of this is already been done, the remaining parts are:
+
+- last_access for usage against timeout value that is ~minutes
+- error_time for usage against retry timeout that is ~minutes
+- uri_worker_map checked for usage against JK_URIMAP_RELOAD=1 minute
+
+So I think, it would suffice to set an actual time at the beginning of
+the request/response cycle (used by everything before the request is being
+sent over the socket) and maybe after the response shows up/ an error occurs
+(for everything else, if there is).
+
+For which cases would it be OK, to use the time before sending to TC:
+- uri_worker_map "checked" (uri map lookup starts early)
+- setting/testing last_access in
+ - jk_ajp_common.c:ajp_connect_to_endpoint()
+ - jk_ajp_common.c:ajp_get_endpoint()
+ - jk_ajp_common.c:ajp_maintain()
+
+What about the others:
+- setting last_access in init should use the actual time in
+ jk_ajp_common.c:ajp_create_endpoint_cache()
+
+- setting last_access again after the service could also use the
+ actual time in jk_ajp_common.c:ajp_done()
+- setting error_time should better use the actual time
+ jk_lb_worker.c service(): rec->s->error_time = time(NULL);
+
+The last two cases could again use the same time, which then would be needed
+to be generated at the end or directly after service.
+
+9) Access/Modification Time in shm
+==================================
+
+a) [Discussion] What will this generally be used for? At the moment,
+only jk_status "uses" it, but it only sets the values, it never asks for them.
+
+b) [Improvement, minor] jk_shm_set_workers_time() implicitly calls
+jk_shm_sync_access_time(), but jk_status does:
+
+ jk_shm_set_workers_time(time(NULL));
+ /* Since we updated the config no need to reload
+ * on the next request
+ */
+ jk_shm_sync_access_time();
+
+two times. So depending on the idea of the functionality of these calls,
+either set_workers_time and sync_access_time should be independently,
+or the second call in jk_status coulkd be removed.
+
+10) "Destroy" functionality
+===========================
+
+[Hint] Destroy on a worker never seems to free shm,
+but I think that was already a flaw without shm.
+
+11) Locks against shm
+=====================
+
+It might be an interesting experiment to implement an improved locking structure.
+It looks like one would need in fact different types of locks.
+In shm we have as read/write information:
+
+Changed only by status worker:
+- redirect, domain, lb_factor, sticky_session, sticky_session_force,
+ recover_wait_time, retries, status (is_disabled, is_stopped).
+
+These changes need some kind of reconfiguration in the threads after
+change and before the next request/response. Since changes are rare,
+here we would be better of, with a simple detect change and copy from
+shm to process procedure. status updates the data in shm and after that
+the time stamp on the shh. Each process checks the time stamp before
+doing a request, and when the time stamp changed it does a writer CS
+lock and updates it's local copy. All threads always do a reader CS
+lock when doing a request/response cycle. Reader CS locks are concurrent,
+writers are exclusive. So readers are not allowed, when the config data is being updated.
+
+Changed by the threads themselves (and via reset by the status worker):
+- counters needed by routing decisions (lb_value, readed, transferred, busy)
+- timers needed by maintenance functions (error_time, servic_time/maintain_time)
+- status is_busy, in_error_state
+- uncritical data with no influence on routing decisions (max_busy, elected, errors,
+ in_recovering)
+
+Here again we could improve by using reader/writer locks. I have a
+tendency for the PESSIMISTIC side of locking, but I think we could
+shrink the code blocks that should be locked. At the monent they are
+pretty big (most of get_most_suitable_worker).
+
+Read-only: name and id.
+
+By the way: at several places we don't check for errors on getting the lock.
+
+12) Global locks
+================
+
+We might want to make the lock technology choosable, like httpd does.
+E.g. on Solaris the default lock type if fcntl, and we can easily
+get an invalid EDEADLOCK for our jk_log_lock.
+
+The following pthread based non global locks are used:
+
+- one mutex for each AJP worker, synchronizing access to the connection
+pool, which exists per process
+
+- one mutex for each lb worker
+
+- a mutex used during dynamic update of uriworkermap.properties to
+prevent concurrent updates. Updates are done per process.
+
+- a mutex to prevent concurrent execution of the process local internal
+maintenance task
+
+- a mutex for access to the shared memory when changing or reading
+configuration parameters. That might be a little unsafe, because it
+actually should be a global mutex, not a process local, but those config
+changes are only done due to interaction with the status worker, so
+there's very little chance for unwanted concurrency here. All dynamic
+runtime data are already marked as being volatile.
+
+All except the last seem to be safe. The last might need some hybrid model
+using thread local mostly and process global when doing updates.
+
+See also: http://marc.info/?t=123394059800003&r=1&w=2
+
+13) Understand the exact behaviour of shm and restarts
+======================================================
+
+Furthermore: rotatelogs (?) and gzip (mod_mime_magic) seem to close
+the (non-existing) shm file. Maybe a problem on some platforms?
+
+14) What I didn't yet check
+===========================
+
+a) Correctness of is_busy handling
+
+b) Correctness of the reset values after reset by status worker
+
+c) What would be the exact behaviour, if shm does not work (memory case).
+ Will this be a critical failure, or will we only experience a
+ degradation in routing decisions.
+
+d) How complete is mod_proxy_ajp/mod_proxy_balancer.
+ Port changes from mod_jk to them.
+
+15) Status worker
+=================
+
+Allow managing pool and connection parameters. Add flags to
+pool and connections to signal workers and maintenance whether
+existing connections should be closed and renewed.
+
+Check completeness of attribute manageability for AJP workers.
+
+Check completeness of runtime data display, e.g.
+reset_time, recover_time, etc. Maybe also add "last error type".
+
+Work on a global display of process local data, e.g. state of the
+process local connection pools (sizes, num connected/idle).
+
+Rework the GUI:
+
+- Basic overview start page with links to the workers,
+ maybe a checkbox to decide whether you want to see config data too.
+- Detail view (what type of error was last and when etc.), detail view
+ for connection pools.
+
+16) URI mapping
+===============
+
+Add more extensions?
+
+17) Connection Pool
+===================
+
+How would a good global maximum count look like?
+Simply limit busyness? Soft limit (applies only to non-sticky requests)
+and a hard limit (applies to all requests)?
+
+18) IP V6
+=========
+
+There's a Bugzilla with a patch.
+
+19) IIS Chunked Encoding
+========================
+
+Move from alternative build to default.
+
+What about the other ifdef'd features?
+
+20) Add REMORE_PORT as a default JkEnvVar
+=========================================
+
+... and port to IIS to fix the getRemotePort() problem.
+
+21) Rework HowTo docs
+=====================
+
+22) Add better example config
+=============================
+
+23) Remove JNI worker
+=====================
+
+24) Watchdog reload of uriworkermap.properties
+==============================================
+
+Questions about uriworkermap.properties watchdog reload (r745898):
+- shm lock needed?
+- Apache port (need to iterate over vhosts)
+
+25) Watchdog backend probing
+============================
+
+Configurable probe URL to test backend, e.g. to decide about
+recovery instead of using random real requests.
+
+26) Commandline shm
+===================
+
+Commandline tool to read data from shm file.
+
+27) Status worker property format
+=================================
+
+Check whether we return list properties as one line
+(because Java doesn't allow the same property key multiple
+times).
+
+Applies possibly to:
+
+"list",
+BALANCE_WORKERS,
+MOUNT_OF_WORKER,
+USER_OF_WORKER,
+GOOD_RATING_OF_WORKER,
+BAD_RATING_OF_WORKER,
+STATUS_FAIL_OF_WORKER,
+