[OAK-2124] Trigger compaction after restart to workaround issue due to long running JCR session

• Type: Improvement
• Status: Resolved
• Priority: Major
• Resolution: Won't Fix
• Component/s: segmentmk
• Labels: compaction gc
• Epic Link: SegmentMK revision GC
• Assignee: Unassigned
• Reporter: Chetan Mehrotra
• Created: 23/Sep/14 14:18
• Updated: 14/Oct/15 09:55
• Resolved: 16/Dec/14 16:24

Description

Due to OAK-2045 we cannot perform efficient compaction on a running system as of now. To workaround that an alternative approach can be used

1. Regular maintainence job (invoked via RevisionGC MBean) if determines that compaction would reclaim space above certain threshold then it sets a boolean flag under "compactionStatus" to true along with timestamp
2. Post restart the maintainence job on detecting such a flag can trigger compaction and cleanup

Or in terms of pseudo code (by tmueller )

// this is called once a day in a background thread
dailyMaintenance() {
    if (notSegmentNodeStore()) {
        return;
    }
    // this needs to be stored somewhere in the repo
    lastCompactStart = loadCompactStartFromRepo();
    if (lastCompactStart > whenThisProcessWasStarted()) {
        return;
    }
    // this can take about 1 minute
    estimatedSpaceSaved = estimatedSpaceSavedByCompaction();
    if (estimatedSpaceSaved < minimumSpaceSaved()) {
        return;
    }
    storeLastCompactStartInRepo(now);
    compact();
}

static long CLASS_LOADED_TIME = System.currentTimeMillis();

whenThisProcessWasStarted() {
    // a simplification, but maybe good enough
    return CLASS_LOADED_TIME;
}

Issue Links

• relates to: OAK-2045 Long running JCR session prevent live cleanup in Segment FileStore