Return-Path: X-Original-To: apmail-ace-commits-archive@www.apache.org Delivered-To: apmail-ace-commits-archive@www.apache.org Received: from mail.apache.org (hermes.apache.org [140.211.11.3]) by minotaur.apache.org (Postfix) with SMTP id 5F79E10A06 for ; Thu, 21 Nov 2013 12:50:44 +0000 (UTC) Received: (qmail 74559 invoked by uid 500); 21 Nov 2013 12:50:43 -0000 Delivered-To: apmail-ace-commits-archive@ace.apache.org Received: (qmail 74151 invoked by uid 500); 21 Nov 2013 12:50:38 -0000 Mailing-List: contact commits-help@ace.apache.org; run by ezmlm Precedence: bulk List-Help: List-Unsubscribe: List-Post: List-Id: Reply-To: dev@ace.apache.org Delivered-To: mailing list commits@ace.apache.org Received: (qmail 74023 invoked by uid 99); 21 Nov 2013 12:50:36 -0000 Received: from arcas.apache.org (HELO arcas.apache.org) (140.211.11.28) by apache.org (qpsmtpd/0.29) with ESMTP; Thu, 21 Nov 2013 12:50:36 +0000 Date: Thu, 21 Nov 2013 12:50:35 +0000 (UTC) From: "J.W. Janssen (JIRA)" To: commits@ace.apache.org Message-ID: In-Reply-To: References: Subject: [jira] [Commented] (ACE-433) New deployment needs two check cycles to download and to install a new DP MIME-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: 7bit X-JIRA-FingerPrint: 30527f35849b9dde25b450d4833f0394 [ https://issues.apache.org/jira/browse/ACE-433?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13828886#comment-13828886 ] J.W. Janssen commented on ACE-433: ---------------------------------- After investigation, I've found out that there are two issues here: # our {{DeploymentServlet}} does not properly send the {{Content-Length}} header in all situations, making the agent unaware of how many bytes to expect; # the {{DefaultController}}, when running in "non-streaming/download-first" mode, starts a download job asynchronously and polls at regular intervals if this download is complete. When it it is finished, chances are that this happens when the controller is not running, so the installation needs to wait until the next time the controller is ran. This is why you need at least two cycles for a DP to be installed. The first finding should be solved in a different issue, which I'll create. The second issue needs some thought, as we need a clear signal that a download is complete to make sure we can install it correctly... > New deployment needs two check cycles to download and to install a new DP > ------------------------------------------------------------------------- > > Key: ACE-433 > URL: https://issues.apache.org/jira/browse/ACE-433 > Project: ACE > Issue Type: Bug > Components: Management Agent > Affects Versions: 1.0.0 > Reporter: Wilfried Sibla > Priority: Critical > Attachments: AgentError.log > > > I tested the new agent implementation but I observed a little bit curious behavior: > * At first check, the agent detects and downloads the update DP > * Then it stops with log entry: [DEBUG] 21:50:31 (downloads) Download stopped early: 926530 of -1 bytes downloaded... (0) > * on the next check: [DEBUG] 21:50:39 (controller) Checking for deployment updates.. > * new DP is detected: [DEBUG] 21:50:39 (controller) Need to install update: newer deployment version available! > * Download is resumed: [INFO] 21:50:39 (controller) Download of deployment update is STOPPED. Resuming download... > * and completed: [DEBUG] 21:51:00 (downloads) Download completed: 926530 bytes downloaded... > This indicates that it needs to cycles to download and really install a DP. > In DownloadCallableImpl.call(): > * is.getContentSize() return -1 during download > * afterwards downloadComplete==false (line 100) and stoppedEarly==true (line 101) indicates that the download seems to be stopped (DownloadState.STOPPED in DownloadResult). > Or is this behaviour (installing a downloaded DP during the next check cycle) intended? -- This message was sent by Atlassian JIRA (v6.1#6144)